Product Details

Digimedia Audioeditor is a professional audio editor with two operating modes: Singletrack for quick editing of a single audio file, and Multitrack for layered, multi-channel recording and mixing sessions. All processing runs locally — no audio is ever uploaded to a server.

Two editing modes

Singletrack fills the canvas with the full waveform of the loaded file — a stereo file shows L on top, R below, separated by a thin divider, with a permanent dB level ruler across the bottom that doubles as a real-time playback meter. Multitrack switches to a lane-based layout: each track occupies its own row with colour-coded clip lanes, and a dockable Mixer panel becomes available with one channel strip per track (Rec / Solo / Mute / Select, fader, peak meter) plus a Master strip.

Non-destructive editing

Unlimited Undo / Redo covers every edit and effect operation — nothing overwrites the original audio until you export. Cut, Copy, Paste, Paste Replace, Delete/Silence and Trim are available from the toolbar or a right-click context menu, alongside zero-crossing selection snapping to avoid clicks at edit points.

Professional effects chain

Every effect opens with a real-time preview and a preset dropdown. The chain covers:

• Gain & Fade — Gain, Fade In, Fade Out.
• EQ & tone shaping — parametric (Paragraphic) EQ, 10- and 20-band graphic EQ with genre presets, and an RNN-based voice noise reducer.
• Compression & limiting — a full-parameter compressor (threshold, knee, ratio, attack, release, makeup gain) with broadcast and mastering presets, a brick-wall hard limiter, and the Normalize tool (see below).
• Time & pitch — time-stretch speed change (0.2×–2.0×, pitch preserved) and combined speed/pitch change with cinematic automation profiles (Doppler, engine, car pass-by).
• Delay & reverb — a feedback delay and a reverb with presets from vocal booth to cathedral hall.
• Distortion & character — saturation/distortion, polarity invert, and reverse.
• Repair & cleanup — three-mode Audio Repair (click reduction, mains-hum reduction with auto-detect, splice smoothing) and Remove Silence.
• Seamless Loop Builder — crossfades a region's tail into its head with adjustable crossfade length and repeat count.

LUFS loudness normalization

The Normalize tool measures Integrated Loudness in LUFS per ITU-R BS.1770 — the method used by every major streaming platform and broadcaster — alongside simpler Peak and RMS modes. A peak-ceiling slider prevents true peaks from exceeding a safe level after adjustment.

For broadcast delivery, EBU R128 sets the reference: −23 LUFS integrated, −1 dBTP true peak, loudness range ≤ 18 LU — also the basis for the AES streaming recommendation and the usual professional podcast target (often relaxed to −16 LUFS for speech-only programmes).

Export formats

Local drafts

Save snapshots to the browser/app's local storage (IndexedDB) without leaving the editor — the whole file, a selection, or the clipboard. Drafts persist indefinitely, work fully offline, and each shows a thumbnail waveform, duration and time since creation.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Cartwall is a broadcast cart wall: a grid of clickable tiles, each holding one audio file, all playing out over a single sound card you choose. Press a tile and the audio starts immediately — built for the split-second timing of a live show. Up to ten tabs organize your carts into sets like Jingles, Beds, Idents and Promos, and up to sixteen carts can play at once. Available for Windows and macOS.

The wall

The grid comes in six sizes — 3×2, 4×4, 5×3, 6×4, 8×4 or 8×8 — from fewer, larger tiles that are easy to hit during a show to denser grids that put a bigger set within reach. Every tile is numbered from 01, counted left to right and row by row, and shows artwork, a remaining-time counter, elapsed time over total length, title and artist, and stereo level meters — tiles adapt gracefully as they shrink on denser grids. A fresh installation starts with ten tabs — Quickcart, Jingles, Beds, Liners, Idents, Sweepers & Bumpers, Effects, Stingers, Promos and DJ Drops — which can be renamed, reordered or closed like any other tab.

Assigning & playing carts

Drop audio files onto the wall from Explorer, or click an empty tile to pick one — mp3, mp2, wav, flac, ogg, m4a and aac are all supported. Cartwall never copies the file: a cart is a reference to wherever the audio already lives. Click anywhere on a tile to start it, click again to stop; up to sixteen carts can play simultaneously, and Stop All silences everything at once. Carts start instantly regardless of length, and an optional Autocue setting skips leading silence so playback begins at the first audible sound.

Editing & organizing

The edit drawer writes artwork, colour, title and artist straight into the audio file itself, so the tags travel with the track wherever it goes. Hold Shift to drag a tab to reorder the strip, or a cart onto another cell to move or swap it. Locked mode hides the edit and delete icons during a show without affecting playback, and removing a cart always asks first — only the assignment is deleted, never the audio file.

Remote control (UDP)

An external automation system can trigger carts over the network: Cartwall listens for UDP commands by default and accepts one plain-text command per datagram.

Configuration

Choose the playout device the wall plays out on, tune Autocue's sensitivity, and set the eight colours the wall cycles through by cell number — a colour saved in a file always wins over the automatic palette. The wall itself lives in library.json, settings in settings.json, both under Micromedia/Digimedia/Cartwall in the platform's application-data folder.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Dashboard fills your screen with a grid of panels: live web pages, RSS feeds, clocks, weather, an audio meter and a notepad, side by side and rearranged by dragging. The news panel can hand its headlines to an AI server on your own network, which rewrites them all in one house style. Available for Windows and macOS.

The dashboard

The window is a dock: a grid of panels, one per widget, each with its own tab above it. A fresh installation opens empty — you add the widgets you want, or load a preset that brings a whole grid with it. Up to 32 widgets can be added under Configuration. The tab of a configured web widget carries a reload button that reloads that page, and tabs cannot be closed: a widget is removed in Configuration, so the grid and the settings can never disagree.

Eight kinds of widget

AI news: one house style for every feed

The News widget reads up to ten feeds at once and drops stories that appear in more than one of them, so the same event never fills the panel twice. Left alone, it shows each feed’s own titles, in each feed’s own voice.

Point it at an AI server — an Ollama server on your own network, entered once under Configuration — and it rewrites every headline in one consistent house style instead. A wire agency, a local paper and a sports feed stop sounding like three different newsrooms, and a presenter can read straight off the screen.

The server is one you run yourself, so the headlines never leave your own network, and it is entirely optional: if none is configured, or it cannot be reached, the widget quietly shows the original headlines rather than breaking.

Arranging and presets

Drag a tab to move its widget: drop it on the edge of another panel to split that space, or onto another panel’s tab strip to stack the two as tabs. Drag the divider between panels to change how the space is shared. Stacked widgets behave like browser tabs — only the active one is visible and the others keep their pages loaded, so switching back is instant. The arrangement is saved automatically. A preset stores the whole grid, every widget with its settings and every panel position; one preset can be starred as the one the app opens with. A loaded preset follows the grid: change a widget and the preset changes with it, so use Save current as new preset when you want to keep the old one.

Ticker and notes

A band of scrolling headlines can be fixed to the top or bottom edge across the whole window. It is not a panel — it has no tab, cannot be dragged, and makes room for itself rather than covering anything. Feed, band height, type size, colours and scrolling speed are all set in Configuration, and if the feed cannot be loaded the band says so rather than quietly standing still. The Notes widget edits one HTML file on your disk with bold, italic, headings and lists; typing is saved back about a second after you stop, and because it is an ordinary HTML file anything else can read it.

Essential information for your podcast or live show

The point of the grid is that everything a presenter needs while recording a podcast or running a live show sits on one screen, with no window to hunt for. What goes on it is your choice:

• Running order and guest notes — a Notes widget holds the script, the questions and the back-timing, edited on air and saved as you type.
• Listener contact — the SMS and MMS console from your messaging provider, webmail, the station’s own internal chat, and the caller management your telephony system offers.
• Social media — Facebook, Instagram and Messenger through their own pages, right next to the microphone, so nothing has to be checked on a phone.
• News, traffic and weather — the News widget merges up to ten feeds into one list with duplicates removed, an RSS widget rotates a single source with its pictures, and the Weather widget shows current conditions and the forecast for the city you are talking about.
• Time — a Clock for your own studio and one for wherever a remote guest is, plus the scrolling headline band along the top or bottom edge.
• Levels — a real PPM, correlation and BS.1770 loudness meter on a sound card input, so the presenter can see the programme feed instead of guessing.
• Video — an HLS or MPEG-TS stream: the studio camera, a return feed, or a news channel running quietly in a corner.
• Audience figures — the dashboard of your streaming or audience-measurement provider, live beside everything else.
• Your own systems — playout, scheduling, the request page, the station intranet: if it has a web interface, it goes on the grid.

Adding one is always the same: add an HTML widget, enter the address, and log in once inside the panel — it comes back ready to use. Zoom the page to fit its panel, switch on automatic reload where a console does not refresh itself, and use the cookie-banner switch to clear the consent prompt these pages tend to open with. When the studio screen should show nothing else, one button puts the whole grid fullscreen.

Your mixing console on the grid

Modern broadcast consoles are controlled over the network, and most of them put that control in a browser. Where they do, the console surface can be a panel like any other page.

Three things to know when a console goes on the grid. Consoles live on the studio network, so the address is usually a plain local one, and the address field takes it as it is. Leave automatic reload off — a reload restarts the page, which on a control surface means dropping out of the session mid-show. And it is live control: a click in that panel moves a real fader on air, so give it a place on the grid where nobody leans on it.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Stacks is a broadcast cart player laid out as a board: between one and sixteen lists sit side by side, each one a fully independent player with its own sound card. A card is one audio file; clicking it arms the cart rather than starting it, so a stray click during a show can never put something on air — Play and Stop are separate, deliberate actions. With up to sixteen lists playing simultaneously, Stacks scales from a single cart set to a full multi-output board. Available for Windows and macOS.

The board

The board holds one to sixteen lists; a fresh installation starts with eight — Jingles, Beds, Liners, Idents, Sweepers & Bumpers, Stingers, Promos and DJ Drops. Every cart in a list shares that list's colour, the chevron in a list header collapses it to a narrow strip (remembered across restarts), and the search box filters cards on screen by title and artist without moving or removing anything.

Arming & playing

Clicking a card arms it — a bright frame appears along with its own Play and Stop buttons — rather than starting playback immediately. Auto select next arms the following cart automatically once the current one stops, by Stop, by reaching the end, or by failing to start. Because each list is its own player, up to sixteen carts can be on air at once, one per list, each through the sound card assigned to that list; while a list has something on air its selection is locked so the running cart can't be swapped out from under itself.

Editing & arranging

The edit drawer writes artwork, colour, title and artist into the audio file itself. Drag a card up or down to reorder it, or across into another list — a playing cart keeps playing while you move it, and only switches to the new list's sound card the next time it starts. Locked mode hides the edit and remove icons during a show, and removing a cart always asks first, naming the title and artist before deleting only the assignment.

Remote control & Control API

An external automation system can drive the board over UDP with one plain-text command per datagram — SELECTA01 arms a cart by list letter and row number, while PLAYA and STOPA always act on whichever cart is currently armed in that list, so the operator's screen and the automation can never point at two different carts. For control surfaces and playout systems, Stacks can also expose the board over a WebSocket Control API — every list appears as one fader, off by default.

Configuration

Each list gets its own name, colour and sound card — or follows the default Playout Device — and the count of lists can be changed from one to sixteen without ever deleting a cart; shrinking the count moves everything from a removed list into the last remaining one. The board lives in library.json, settings in settings.json, both under Micromedia/Digimedia/Stacks in the platform's application-data folder.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Explorer is a free desktop application for broadcast engineers and audio professionals. It automatically discovers every Audio-over-IP stream on your network, lets you pre-listen to any stream on your local sound card, route audio between Dante and Livewire devices, and provides broadcast-grade metering — all in a single window, with no configuration required. Available for Windows and macOS.

What it discovers

Explorer passively finds and lists every stream and clock on the network, grouped by discovery mechanism: AES67 (SAP) senders announced via SAP/SDP, and mDNS — which covers RAVENNA, Dante and Axia/Livewire in one unified list. Clock discovery covers PTPv1 and PTPv2 grandmasters. It also finds network devices advertising a web management interface (DHD, Digimedia, Axia and more) and opens them in one click.

Key capabilities

• Passive, zero-configuration discovery — Explorer never alters your network or joins a multicast group in a way that affects the sender.
• Pre-listen to any stream on any attached sound card, with selectable channel mapping for multichannel sources.
• Broadcast-grade metering: stereo PPM (IEC 60268-10), ITU-R BS.1770 LUFS, Lissajous goniometer, phase correlation and a scrolling peak-level history.
• Route audio to Dante and Livewire receivers — drag-and-drop per receiver, or a full routing matrix in the Router tab.
• Per-NIC filtering so you can focus on one network segment at a time; discovery still runs on all NICs simultaneously.
• 17 colour themes, dark and light mode, five languages.
• Built-in update channel — one click to download, verify (SHA-256) and install new releases.

Metering panel

While a stream is monitored, the right-hand rail shows three stacked views. The goniometer is a green-phosphor Lissajous plot of the stereo field — circular for mono-compatible content, a horizontal ellipse for out-of-phase content, a vertical line for identical L/R. The PPM tab shows stereo peak programme meters with IEC ballistics and a phase-correlation strip. The LUFS tab shows ITU-R BS.1770-4 momentary loudness with a running maximum. A scrolling level-history bar chart is pinned at the bottom, colour-coded from blue (≤ −18 dBFS) through green, orange and red (> −3 dBFS).

Clock & device discovery

The Clocks tab lists every PTP grandmaster on the network with full quality detail — priority values, clock class, accuracy, steps removed and time source for PTPv2. Explorer only listens; it never participates in the BMCA election. The Devices tab lists devices advertising an HTTP/HTTPS management interface, filtered by a configurable MAC-OUI allow-list (Micromedia AG and dhd.audio are included by default).

Routing to Dante & Livewire

The Receivers tab lists Dante devices with at least one RX channel and Livewire destinations — AES67 and RAVENNA never announce a receiver identity on the network, so they aren't offered as routing targets. Drag a compatible source from the Compatible Streams panel onto a receiver channel to route it: a Dante source subscribes via Dante's own ARC mechanism, while an AES67, RAVENNA or Livewire source binds directly to its resolved multicast address once available. Livewire destinations show a single "Channels 1+2" row and accept a source of any protocol, since Livewire's own routing protocol only needs a multicast address and port.

For a full overview, the Router tab shows every transmitter and receiver at once as a matrix — click a cell to subscribe or unsubscribe a route, or expand a row or column to route individual channels. A yellow spinner marks a route in progress, a green checkmark a confirmed route, and an amber warning icon a Dante subscription reporting a fault; Clear unsubscribes every active Dante route at once.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Surface turns your computer into a mixing control surface for any DHD audio core. Pick one of three connection modes per workspace and ride faders, ON and PFL on a familiar console strip. Drive several cores at once in separate tabs.

Three connection modes

Each surface uses exactly one mode, chosen from a drop-down on the connection bar.

dhd.audio API and Ember+ both control a DHD mixing core and render the same strip layout. MIDI MCU is the opposite role — the DAW controls Surface, which emulates a Mackie Control surface.

The mixer console

Each channel strip carries a source drop-down (pick which core fader the strip controls), a PFL / headphone button, a fader with a dB scale drawn on every strip, a stereo PPM level meter (dhd.audio mode only — Ember+ exposes no metering), and an ON button. Show 1–24 strips at once via the Faders slider; hold Shift and drag to reorder a strip, or click Reset Layout to restore the default source mapping.

MIDI MCU mode

In MIDI MCU mode, Surface emulates a Mackie Control surface that your DAW drives over RTP-MIDI. Each strip carries the fader plus four DAW-lit buttons — Rec, Solo, Mute and Select — with channel names, time-code and meters pushed live by the DAW. Up to 32 parallel MCU connections can run at once, one per sub-tab.

Multiple surfaces

Add as many surfaces as needed — mix dhd.audio, Ember+ and MCU surfaces freely in separate tabs. Each keeps its own host, mode, layout and connection state, remembered between sessions.

Resilient connection

The dhd.audio link is kept warm with a periodic keep-alive and reconnects automatically if the core drops or freezes. On disconnect the surface goes idle — faders fall to −∞, ON/PFL clear and meters empty — so nothing on screen reads stale; current positions and states are read back automatically on reconnect.

Appearance & updates

17 colour themes in dark and light mode, five UI languages. Surface checks for a newer release at startup and via the bottom-bar link; updates are SHA-256-verified before install, with independent Windows and macOS release tracks.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Sounds puts up to 99 Sonos™ zones on one screen for studio and broadcast use — each zone gets its own card with now-playing information, transport controls, a volume fader and a browser for favourites and playlists. Nothing is installed on the speakers: the app talks to them directly over your own network. Digimedia Sounds is an independent third-party application; Sonos is a trademark of Sonos, Inc., and Micromedia AG is not affiliated with, endorsed by or sponsored by Sonos, Inc. Available for Windows and macOS. Digimedia Sounds controls Sonos™ hardware and does not play audio itself: at least one Sonos™ hardware player, such as the Sonos™ Port, is required.

Cards and zones

Open ⋮ → Configuration to set up between 1 and 99 players — each one a card, each pointing at a Sonos zone found on the network. A single reachable speaker reports the whole household, so setup is quick even on a large system. The window is a single row of full-height cards that share the width evenly and never wrap; the strip scrolls sideways once the cards no longer fit. A dot beside each card's name is green while its zone is answering, red while it isn't, and grey when nothing is assigned; Use zone names re-titles every card after the zone it points at.

Loading and playing

Every card follows one rule: picking something never puts sound in a room. Clicking an entry in a card's browser only loads it — the card shows the take and its badge reads Loaded; the ▶ button is what actually starts it. If a zone is already playing something else when you load a new take, the card adds a green ▶ line naming what's actually on air, so it never implies a take is playing when it isn't. The loaded take is remembered between sessions, so a card comes back armed the next morning.

Transport, volume & My Sonos

Each card carries play/pause, stop, a draggable progress bar, mute and a volume fader for that zone's own level — a live radio stream has no length, so its bar stays empty. The lower half of every card is a browser with three tabs: Favourites (everything in My Sonos, stations included), Playlists, and Search, which matches favourite names, playlist names and the tracks inside playlists. Cards refresh from the speakers a little under twice a second, so a change made from the Sonos app shows up here on its own.

Playlist behaviour & grouped zones

Two switches in Configuration → Playlist decide what a loaded playlist does: Automation makes it repeat with crossfade instead of stopping after one track, and Randomize shuffles it — only when Automation is also on. Both are sent to the zone every time a take starts, so a shuffle never outlives the setting that asked for it. Sonos only accepts transport commands on the zone leading a group; a card pointed at a follower has no controls worth showing and collapses to a narrow spine instead, expanding again the moment it's made independent.

Remote control

An automation system can drive the cards over UDP with one plain-text datagram per command — each command carries a two-digit card number from 01 to 99, counting the cards from the left, on by default. For control surfaces and playout systems, Digimedia Sounds can also expose every card as a fader over a WebSocket Control API, off by default — a switch remaps the top of the fader between 0 dB and +10 dB, and the app paces what a motor fader sees so it settles onto a value instead of hunting between steps.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Signage lets you lay out a BrightSign digital signage screen visually — a background or overlay, HTML and RSS widgets, and up to two video windows — on a canvas that mirrors your target resolution exactly, then deploy the finished layout straight to a real BrightSign player over the network. There's no BrightScript to write by hand: the app compiles the layout and uploads it for you. It's equally at home driving a broadcast studio screen — showing a live dhd.audio Views App view — as a lobby or retail display. Available for Windows and macOS. Digimedia Signage designs and deploys the layout, while playback runs on the player itself: a BrightSign™ hardware player is required.

What a layout is made of

Every design is built from five kinds of thing, all placed on one shared canvas: one background (a solid color, an image, or a live web page), an optional overlay (an image or live web page shown on top of everything else — a logo, watermark or lower-third — mutually exclusive with the background), any number of HTML widgets (a web page shown in a positioned rectangle), any number of RSS widgets (a rotating news-style card fed from an RSS or Atom feed), and up to two video windows. With a background active, video always renders in front of every widget; switching to an overlay instead moves video behind everything, so the overlay and your widgets can sit on top of it — matching how a deployed BrightSign player composites these layers. Everything is saved automatically as you work and restored the next time the app starts.

Project Settings fix the target screen's Resolution, Orientation (horizontal, or vertical flipped 90° or 270° for the physical mount) and Frequency — independent of each other, so rotating the mount doesn't change which native resolution was picked.

HTML & RSS widgets

An HTML widget shows any web page inside a rectangle you position and size on the canvas; a Suppress Popups option sets a standard user agent and hides known cookie-consent overlays (OneTrust, Cookiebot and others) on a best-effort basis. An RSS widget fetches an RSS or Atom feed and rotates through its items — headline, lead text, an image if the feed provides one, and the feed's own source name — refreshing automatically every time it completes a full rotation. Picture placement (None, Left, Top, Right or Background) and independent title/lead text sizes are configurable per widget.

The HTML widget is also what makes Digimedia Signage a natural fit for a broadcast Studio Screen: point it at a dhd.audio Views App URL in standalone mode (http://[core-ip]/#/views/[id]?standalone=1) to show a live console view — clock, meters, routing or any other configured view — full-screen on a real player, with no separate PC or browser kiosk needed.

Video windows & live preview

Up to two simultaneous video windows play an rtsp:// or https:// stream, scaled to a locked 4:3, 16:9 or 21:9 ratio, or resized freely. Preview Content — available on every widget and video window — plays the real page, feed or stream live inside its canvas rectangle while designing, as a design-time convenience only: it is never part of what gets deployed, and because the designer and the deployed player use different network paths and rendering engines, a failed preview does not necessarily mean a failed deployment (or vice versa).

Deploying to a player

The Players drawer keeps a list of BrightSign players, each identified by name, IP address and its BrightSign Local DWS username/password (default username admin, with the device's own serial number as the password unless it has been changed). Clicking the upload icon on a player's row connects to it, reads back its device info, compiles the design to BrightScript, uploads the background image and the compiled script, then reboots the player straight into the new layout — all in one click, with a status pill (Idle, Uploading, Uploaded or Error) tracking progress on screen.

Projects, appearance & updates

A design — layout and background image together — packs into a single portable .sig file via Save, independent of the player list, and reopens with Load. Several color themes are available in both light and dark mode. A built-in updater checks for a newer release, then downloads, verifies and installs it before relaunching — Windows and macOS are updated from separate channels.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia MIDI bridges DHD (dhd.audio) consoles and Digimedia applications — Playout, Stacks, Communicator, Cartwall, Audioeditor and Sounds — to an iCON fader-surface controller over USB-MIDI, with free per-channel mapping between them. It runs as a background Windows service and is controlled entirely through a browser-based web interface, with no internet connection required.

How it fits together

Every Digimedia application — and, if you have one, a dhd.audio RM1 console — connects to the bridge as a Core. The bridge pools all of their channels and maps that pool freely onto the connected iCON P1-X surfaces, so any fader strip can carry any channel from any source.

Digimedia applications & DHD cores

Add any number of Digimedia applications and DHD consoles as Cores, each on its own sub-tab. A Digimedia application — Playout, Stacks, Communicator, Cartwall, Audioeditor or Sounds — connects to its own Control API server by address and token; a DHD console connects over its WebSocket API. Either way the sync is real-time: faders, meters and buttons move live in both directions. A status indicator shows each connection at a glance: green (connected), amber (scanning) or red (disconnected).

iCON surface support

Supports the iCON P1-X and V1-X extenders over USB-MIDI — plug and play, no configuration required. Digimedia drives the extenders directly: a master unit is neither required nor supported. Up to 7 surfaces can run simultaneously, covering up to 56 faders, and the V-Pot above each fader works on both models. Surfaces with motor faders are recommended, so that a level changed anywhere else follows on the surface.

Channel mapper

The Mapper is a matrix-style editor: core nodes on the top expose the output handles, surface nodes sit on the left. Just click a connection point to wire up bidirectional sync — fader moves on the console move the physical motors on the surface, and vice versa, synchronising within seconds.

V-Pot channel browsing

A surface has eight faders while an application can offer many more channels, so mark the ones you use often as favourites — click a channel name in the Mapper and a star appears. The rotary V-Pot above each fader browses exactly those: rotate to scroll through the favourites on the LCD, press to move the selected channel into that fader strip. A channel that already sits on another fader is moved there rather than duplicated. Browse mode times out after 5 seconds of inactivity, cancelling the selection.

Control API server

Besides driving a fader surface, Digimedia MIDI can hand everything it is connected to on to another program. Its built-in Control API server presents every connected application and console as one single aggregated mixer, so a client such as Digimedia Surface can read and set every exposed channel — fader level and on-air state — from a single connection, with live level and metering updates. Digimedia Surface can connect in two ways: as an MCU surface over RTP-MIDI, which is fixed at eight faders, or over the Control API, where the number of faders is free. Only the channels starred as favourites in the Mapper are exposed, so you decide what leaves the bridge.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

Digimedia Communicator is a multiline, multi-fader talkshow phone software built for any size of radio studio. Every list is a place a call can be, every card is a live call, and every fader is its own stereo conference room on its own sound card — drag a caller onto a fader to put them on air. Each room can be recorded as two files: the callers alone, and the full exchange with the studio. The Recordings list beside the board plays these files back, and the built-in Audioeditor edits them while the phone stays on air. Available for Windows and macOS.

The board: Line, Hold, faders and recordings

Left to right: Line, Hold, then up to sixteen faders, with the Recordings list at the far right. Line is where every incoming call arrives and where an outgoing one starts unless it was dialled from a fader's own dialler — you hear and answer it on the Line sound card (your screening headset), and two callers waiting there can't hear each other, by design. Hold parks a call and feeds it whatever arrives on the Hold sound card, your mixing desk's hold music or programme output; nothing comes back from a held caller. Each fader is a conference room: everyone in it hears everyone else and that room's own sound card, and the room hears them back — rename rooms and choose their sound cards in Configuration. A station that only screens and parks callers can run with no faders at all. The phone runs by itself: the SIP engine starts with the app and stays up, and if it ever cannot start, a message above the board says so and stays there.

Moving a call

Dragging a card is the whole application. Line → Hold puts a call on hold; Hold → Line brings it back to the screening bus; Line → a fader answers a still-ringing call first and moves it into that conference room, since dragging a caller somewhere they can be heard is an unambiguous instruction to take the call; fader → fader re-routes a call live; fader → Hold parks it and takes it off that room's bus. A card only moves once the phone software confirms it — if it snaps back, a toast explains why. Taking a call off hold asks the far end to resume, so a card can stay amber for a moment before turning green; dropping a ringing call on Hold does not answer it.

A card shows the caller's name or number, the account and the codec, a timer that starts when the call connects, and two meters: the caller on the left, what the room sends them on the right. Its colour is its state — green while connected, amber while held, blinking between green and amber while it rings, grey once it has ended. Hang up always sits on the right, so the one control whose action cannot be undone never moves. Clicking the body of a card does nothing on purpose: during a phone-in, a stray click must not be able to put anyone on air or drop them.

Calling out, transfer & digits

A + in a list header opens the dialer aimed at that room — pick an account, type or click a number, and press Call or Enter. A call dialled from a fader's own + appears directly in that fader from the first moment, already ringing out, without ever passing through Line. Find a contact searches by name or company as you type; clicking a result only fills in the number, so you can check you have the right person before dialling. Digits opens a keypad that sends DTMF the instant each key is pressed. Transfer either hands a caller to a typed number and drops out immediately, or bridges them with another connected call and drops out of both.

Lines, codecs & encryption

Up to eight SIP accounts can be registered at once, each on its own tab in Configuration with a dot that shows its state at a glance — green registered, red failed, amber still trying — so a line that is down is visible without opening them all. Each account chooses its transport, UDP, TCP or TLS, and its media encryption: Off; Offer, which encrypts the audio whenever the far end supports it and otherwise falls back to unencrypted audio; or Require, which refuses a call that cannot be encrypted. Encrypted audio is only as safe as the signalling that carries its keys, so Configuration warns you when encryption is on without TLS. Codecs are a ticked list in order of preference — Opus and G.722 first for wideband, G.711 underneath so that no provider is unreachable — and each account can carry its own ring tone, so two lines can be told apart by ear. Passwords are stored encrypted and tied to the machine, and everything runs at 48 kHz stereo.

The mixer & recording

The Mixer button in the bar along the bottom of the window opens a row of strips below the board, pushing the board up rather than covering it. The leftmost strip, called Tel Rec unless you rename it, sets how loud recordings play back; it is not a conference room, so a recording can never be played onto a fader that is on air. The strips to its right are the rooms, in board order: drag one to set that room's output level, and double-click it to return it to 0 dB. Line and Hold have no strip, since neither has an output level to set. The levels are remembered across restarts.

Recording is started from the record dot in a list's header: Line and every fader have one, while Hold, a one-way feed, has nothing to record. Every take produces two files at the same time: one with the callers alone, and one that adds the studio end from the room's own sound card. The caller-only file stays usable on air whatever the studio microphone picked up — a cough, talkback or headphone bleed — while the mix keeps the whole exchange; a single file would force that choice before anyone knows which one is needed. Every card in a recording room shows a red rule, because stopping the recording stops it for all of them at once. A caller dragged into a room that is already recording joins the take automatically, and a room whose sound card is missing can still be recorded — its callers are captured all the same.

Recordings & the Audioeditor

The Recordings list to the right of the board shows every take in the recording folder, newest first, with its room, date, length and size, and Play, Pause, Stop, Edit and Delete on each card. Playback goes out of a sound card you choose in Configuration, not the system default — which on a studio machine is often a card nobody has plugged anything into. Selecting a take only arms playback, which cannot reach anything on air; the selected take is also what the Control API's recordings fader plays.

Edit raises the built-in Audioeditor from the bottom of the window while the board keeps running underneath: calls still arrive and recordings still start and stop, because an editor over a live phone board must never take the phone off the air. It is a single-track editor with gain, normalisation, compression, limiting and fades, equalisers, noise reduction tuned for speech, silence removal, speed changes and more, and it works on a copy in memory until you save. Save as copy exports in the format and bit depth you choose. Overwrite asks first, because a phone-in cannot be recorded again; the original stays whole until the replacement is completely written.

Contacts, history & remote control

Contacts holds names, numbers, companies and notes, with a favourites filter and one-click dialling from any row; the search covers the whole address book, not just what's on screen, and stays fast on a large one. Call History lists every call with its time, length and account, filterable by Incoming, Outgoing or Missed, and any row can be called back or added to Contacts. An optional Control API lets a control surface or studio automation system drive the faders over the network and read back their state: each room's level and meter, how many callers are in it, who the first of them is, and whether it is being recorded. Switching a room's fader on takes the longest-waiting caller out of Line and into that room; switching it off parks that room's callers on Hold rather than hanging up, so a knocked key can never lose a caller. In the Control API, the first fader plays the recording selected in the Recordings list, and the rooms come after it. Watching needs nothing, every change needs a token, and the API stays off until you switch it on.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia VAD is a Windows kernel-mode virtual audio driver and control service that turns a standard PC into a professional AES67 / RAVENNA audio-over-IP node. The driver appears in Windows as standard audio endpoints, so any application can play or record network audio with no special software — configured live from a built-in web control panel.

Architecture

Two parts run together on the host PC: a kernel driver (WDM/PortCls) that transmits and receives AES67 RTP packets — up to 8 render and 8 capture instances — and a control service hosting an HTTP/WebSocket server that configures the driver and announces active streams via SAP, mDNS and RTSP. The web control panel is served directly by the service, reachable from any browser.

AES67 transmission & reception

Each of the up to 8 independent instances pairs a render endpoint (TX) with a capture endpoint (RX). TX streams are configured with multicast group, port, sample rate (48/96/192 kHz), packet time, DSCP QoS marking and payload type; the SAP and mDNS/RTSP advertisements update within a second of applying changes. RX streams discover network sources passively via SAP and mDNS, with drag-and-drop assignment from the discovery list. Both directions support SMPTE ST 2022-7 seamless redundancy over a secondary multicast group and NIC.

Dual-slave PTP with automatic failover

Each instance runs a software IEEE 1588-2008 (PTPv2) slave entirely in the Windows kernel. Two independent slaves (PTP 1 / PTP 2) can track grandmasters on separate NICs; a five-level priority cascade selects which one drives the media clock, with LOCKED always beating UNLOCKED — so if the primary NIC loses its grandmaster, an already-locked secondary slave takes over instantly with no clock jump.

Web control panel & API

The Dashboard gives a live overview of both PTP slaves and all RX/TX flow states. All functionality is also available over a WebSocket Control API — including a DHD-compatible surface that lets existing DHD ecosystem tooling enumerate and route streams without modification. Remote access can be protected independently with a UI login password and an API bearer token; connections from the machine itself are always exempt.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

Digimedia Playout is a playlist-driven on-air system for radio broadcast automation. The rundown is the running order — pull a fader up to start the Next element, double-click a row to start it immediately, and up to five independent output channels run their own players in parallel. An integrated OnAir TrackMixer handles the transitions between tracks. Available for Windows and macOS.

Rundown, Next and players

The rundown is the running order — there is no separate queue, so moving an element is how you change what plays next. The Next position is the top row: the element that starts on the next start, always something with audio. Below it, Info rows carry commands and notes with no audio of their own. Pull a fader up or press ON to start Next on that channel, use Start Next to reuse the last channel, or double-click any row to start it immediately. The element started last always takes the rightmost player; when none is free, the one nearest its end is stopped to make room. Stop something within five seconds of starting it by mistake and it returns to Next without counting as broadcast.

Operating modes

Manual leaves every start to the operator; Assist runs segues automatically while blocks stay manual; Auto runs the rundown unattended; Standby halts playout while keeping the rundown loaded. Up to five independent output channels each run their own player, so several rundown positions can be on air with different destinations at once.

The players

Each player carries two time read-outs, switchable by right-click between Elapsed, Remaining and Total. Remaining counts down to Start Next rather than the end of the file — the moment that actually matters on air — and only switches to the true end afterwards. While an intro plays, the second read-out shows its countdown regardless of what it's set to. Double-clicking a player's progress bar opens that transition directly in the OnAir TrackMixer.

Editing the rundown & marks

Drop a track between two rows to insert or move it, onto a row to swap it, drag it in from the library to schedule it, or double-click a library track to schedule it straight after Next. Deleted elements go to a recycle view, draggable back at any time. Marks — Mark In/Out, Fade In/Out, Intro, Outro, Start Next, Link In/Out and Hook Start/End — live on the track in the library, not on the rundown element, so they travel with it into every future broadcast.

OnAir TrackMixer

The transition editor holds up to nine tracks on one shared timeline, already assembled whenever Link Out and Link In are set. Drag a track along the timeline to change the overlap, add or remove level points on its envelope, and use Duck to lower every other overlapping track as editable level points, or Crossfade to build the transition and set both Link marks automatically. The region before the first track's Mark In and after the last one's Mark Out stays locked, since it hands over to an element that isn't loaded yet.

Configuration

Set a name and sound card per output channel, with Auto Cue to skip leading silence on tracks that have no Mark In of their own. Choose between two and five players, and point the library at a folder — re-scanning it keeps every mark already set. Changes apply immediately; there's no separate save step.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Logger records what you broadcast, around the clock, so you can prove it later. It runs unattended as a Windows background service, keeps up to eight audio and video channels rolling in one-hour segments, and watches every one of them for silence. The whole interface is a browser page: no client to install, an archive you can scrub through, and a trimmed clip exported in seconds.

What it records

Each instance carries up to eight independent channels, and a channel takes its audio either from a physical soundcard or from a network stream — RTSP, RTMP or HTTP/HLS. Stream channels can capture the parallel video alongside the audio.

Recording is continuous rather than file-by-file: audio rolls into one-hour segments without gaps, and segments older than the channel's retention setting are pruned automatically, so a logger set to ninety days keeps ninety days and never fills the disk.

The live view

The Recording workspace gives every channel its own strip. Standby switches the channel's engine on or off, so a channel you are not using costs nothing; REC arms and disarms the recording itself and pulses red while it runs.

Monitor streams that channel's live audio to your own browser for a confidence check, without touching what is being recorded. On stream channels a single button flips the strip between a live video preview and the goniometer.

Metering

Two standards run side by side, switched by a tab on the strip. The peak programme meter shows both channels in dBFS with colour-coded caution and danger zones, and a phase correlation bar underneath tells you at a glance whether the signal will survive a mono downmix.

The EBU R128 meter reads momentary loudness on a 400 ms window and gated integrated loudness over time, with its own elapsed timer and a reset button so you can measure a single programme rather than the whole day. The goniometer draws the stereo image as a polar scope: a vertical line is mono, a round cloud is wide.

Silence detection and alerts

Every channel is watched for signal loss. Fall below the configured threshold for longer than the configured window and the strip frames itself in pulsing red, a banner names the channel and counts the silence, a webhook fires at whatever supervision system you run, and an email goes out over SMTP.

Recovery is announced the same way: when the signal comes back, a cleared notification follows on both channels, so nobody is left wondering whether the fault is still live.

Archive, trimming and export

The Archive workspace opens on a calendar and a grid of the day's twenty-four hours, each marked with what exists: audio, audio plus video, or nothing. Picking an hour loads its waveform from a pre-rendered peak cache, so it appears immediately rather than after a decode.

Set the in-point and the out-point by dragging, or clamp either one to the playhead, audition just the selection, then export. Audio comes out as a lossless WAV cut from the original; video is remuxed server-side into an MP4 with its audio in sync. Nothing in the archive is altered by any of it.

Configuration, storage and access

Each channel carries its own name and colour, source, stereo or mono, sample rate, encoding — lossless PCM for archival, or MP3, OGG or MP2 where storage matters more — recording path and retention in days. The configuration dialog includes a folder browser for picking a dedicated drive, and the manual gives storage-per-hour figures per format so a retention policy can be planned rather than guessed.

Access is deliberately two-tier. A browser on the logger machine itself always gets in without a password, so an engineer can never be locked out of the box in front of them. Anything coming from the network is refused outright until an administrator password is set, and is then challenged for it; passwords are stored salted and hashed.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Dynamics is the processor that sits in the transmission path. It takes a stereo signal from a soundcard, runs it through twenty-five stages — levelling it, shaping its spectrum and holding an absolute ceiling — and returns it to a soundcard. Alongside that it feeds up to eight internet streams and records what went out in hourly files, and eight slots take your own VST plugins wherever the chain needs them. Dynamics runs as a background service, so it starts with the machine and keeps going whether or not anyone is looking at it; the interface is a browser page and only a remote control.

The chain

Every stage is a row, in signal order, showing its number, an ON or BYP button, and a one-line summary of what it is currently set to — plus a small live meter where there is something to show, such as gain reduction. Click a row to open that stage; click ON or BYP to take a single stage out of the path without disturbing its settings. Bypassing removes a stage from the signal path entirely rather than merely zeroing its effect, and the Bypass button in the header does that to the whole chain at once, which is the honest A/B against the untouched signal.

Most of the chain can be reordered by dragging a row. The first four stages are fixed at the front because they work on the converter's own waveform and have to see it unaltered, and the last two are fixed at the back, because a true-peak ceiling is only a guarantee if nothing follows it. Bass-Mono has no handle of its own: it rides directly behind the Stereo Enhancer, whose widening it exists to undo in the low end. The order lives on the server, so every connected browser sees the same chain.

What the stages do

The chain reads front to back as a sequence of jobs. Work from the front: a problem fixed early does not have to be fought later, and every stage left bypassed is one less thing colouring the sound. Eight slots for your own plugins can also sit between these stages; those slots are described in the next section.

Your own plugins

Eight slots take any 64-bit VST3 or VST2 plugin and put it straight into the broadcast chain. Each slot is a row like any stage, so you drag it to wherever it is needed — a favourite de-esser early on, a tape emulation ahead of the multiband compressor, a stereo widener just before the limiter. The fixed head and tail stay where they are, so nothing can end up behind the ceiling. Every slot has its own wet/dry mix for parallel processing, and its own bypass.

A plugin that crashes does not take the broadcast with it: its slot is switched off, the incident is logged, and the audio keeps running.

The workspace

Three columns: the meters on the left, the chain in the middle, and the parameters of whichever stage you selected filling the rest. Every detail panel is built the same way — the picture first, then the live values, then the controls — so once you have learned one stage you can read all of them. A question-mark button adds a description of the stage above the picture, and it is one switch for every panel: leave it on while you are learning the chain, off once the pictures speak for themselves. A reading with no signal behind it shows a dash rather than a number.

Several panels let you work in the picture instead of the sliders: drag an EQ band sideways for frequency and up or down for gain, drag a compressor's threshold line, drag the crossovers between bands, drag along the stereo width arc. The sliders follow, and they stay the authority on range and step size. Every slider takes a typed value, steps with the arrow keys, goes back to its default on a double-click, and moves ten times finer with Shift held.

One warning worth reading twice. This processor has no monitor path — its output is the transmission. So the Solo button in a multiband band editor mutes the other bands for every listener, not only for you. While a solo is live, every connected browser shows a red notice above the chain with a button to clear it, and any browser may clear it.

Metering

The meters column exists so you can compare the input with the output: two stereo meters side by side, with a switch between peak-programme and loudness. The PPM bars follow IEC 60268-10 ballistics — a fast rise, a decay of about 11.7 dB per second and a peak-hold line — while the loudness reading is EBU R128, momentary over a 400 ms window for what the ear is hearing now and gated integrated for the number a regulator cares about. A true-peak readout in dBTP sits alongside.

Above them, a goniometer draws the output as a Lissajous figure and a correlation bar says the same thing as a number: vertical and +1 is mono, a wide horizontal cloud below zero is partly out of phase. Below −60 dBFS the correlation glides to the centre and reads exactly zero, because a needle swinging across a third of the scale on an idle machine reads as a fault.

Ten of the panels draw a one-sixth-octave spectrum behind their curve: the input as a faint area, the output as a line. Showing both is the point — the gap between them is what the chain is doing to the material at that frequency.

Presets and banks

Seven presets ship with the processor. Signature is the house sound and the default: more bass and punch than the source, with the top end as the recording has it. Powerful is warm, dense and pushing; Broadcast Music is the same idea held back; Speech is built for the spoken word; Classic lets large dynamics survive, so a pianissimo stays quiet; Stream is dense, built for the loudness normalisation of streaming platforms, with the most true-peak reserve of all; and Transparent keeps only the protection — the Bandpass high-pass and the true-peak limiter — and so serves as the honest comparison point for the others. The louder presets deliberately leave more true-peak headroom, because a lossy encoder adds overshoot of its own. Store what is running under a name of your own at any time; a preset you save is a complete snapshot, including the order of the chain.

Three banks sit under the chain header: Chain A and Chain B, the two production banks, and Temp, a scratchpad. Copying a bank into Temp also switches the output to Temp, so what you try there is on air while the original bank stays untouched; once you like the result, commit Temp to either production bank. A switch between banks lands on a buffer boundary and is blended across that boundary, so a change on air is inaudible. There is no level matching anywhere in the product: what you hear is what goes out, so to compare two banks fairly by ear, match their loudness targets first.

Straight to your listeners

Dynamics carries its own streaming encoders, so the processed signal goes to your internet radio servers without separate encoder software. There are eight independent connections, each switched on and watched on its own: Icecast 2 with MP3, AAC, Ogg Vorbis or Opus, and Shoutcast v1 and v2 with MP3 or AAC, at 64 to 320 kbps. A web player can take a high-quality stream while mobile listeners get a lighter one, both from the same processing. Each connection shows the bitrate it is actually delivering, and says plainly when it is connected but has no audio instead of pretending to be live.

Track titles reach MP3, AAC and Ogg Vorbis listeners as the songs change. Your playout system can write them to a file, send them as a UDP datagram or pass them over WebSocket; Dynamics takes each change to every connected server, and a stream that reconnects is sent the current title at once, so nobody is left on a song that has already ended.

An archive of what went out

The File Writer records the processed output continuously as 24-bit stereo WAV — one folder per day, one file per hour — in exactly the format Digimedia Logger reads, with a peak file beside each recording so that Logger can scrub it without reading the audio. A gap of up to two seconds stays in the same file; after a longer silence the file is closed and the audio that follows starts a new one. An existing file is never overwritten, not even when the clock repeats an hour at the end of daylight saving time. The destination is chosen with a built-in folder browser that lists the drives of the machine running the service, not those of the one you happen to be browsing from. Recording can never interrupt the broadcast: a slow or full disk costs you the recording, never the air.

Running it unattended

The processor is built to sit in the transmission path for months. Every parameter change takes effect live, without interrupting audio. Applying a preset, reordering the chain or reloading a stage is recorded with the time and the user, so the question of who changed the sound, and when, has an answer. Changing a soundcard or the buffer size is the one kind of action that restarts the audio stream and costs a brief dropout, which is why both live in the configuration rather than on the main screen. Input and output can be two different cards. Their two clocks then drift slowly apart, and drift compensation keeps that from causing a click every few minutes.

Access is two-tier. A browser on the machine itself is never asked for a password, which is what makes it impossible to lock yourself out, and access from the network is refused outright until a password has been set — not merely unprotected. The password can only be set locally, so nobody on the network can claim the first one. Updates are offered in the About drawer rather than fetched on a schedule: the installer's checksum is verified before it runs, and your configuration, presets and change trail live outside the installation directory, so they survive every update.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Deepfilter cleans up speech and then finishes it. Each instance takes a signal from one soundcard, runs it through a DeepFilterNet neural pass that removes what is around the voice, sends it on through a nine-stage broadcast chain, and returns it to another soundcard. Up to eight instances run side by side on one machine, each with its own pair of cards, its own model and its own settings — a presenter, a guest and a telephone hybrid, tuned separately. There is no desktop window: it runs as a background service and the interface is a browser page. Beta phase running.

One Deepfilter per soundcard

An instance owns one input card and one output card and nothing else. The card's first two channels are summed to mono on the way in, so a source wired to the left leg, the right leg or both always arrives as a single signal; the finished result is written back to both output channels.

That makes the natural unit one microphone path per interface. A presenter, a guest and a telephone hybrid can sit on three cards in the same machine, each with its own noise model and its own chain settings, and none of them touches the others. Sample rate, buffer size and audio system are shared across the machine; everything else belongs to the instance.

Each instance also opens its own two cards. One that is unplugged, or held exclusively by another application, stops that instance alone and says so on its tab — the rest keep streaming. Instances are added and removed under Audio & model, from one to eight; removing one keeps its settings, so raising the count again brings it back as it was.

The chain

Every instance runs the same nine stages in the same order, and each one can be switched out on its own with its BYP button — the quickest way to hear what a stage is actually contributing.

The AGC is deliberately slow: it evens out speakers who work at different distances and levels, it is not a second compressor, and how far it may go in either direction is bounded. If Brickwall TP shows constant gain reduction, the level arriving at it is too high — lower it upstream rather than leaning on the limiter.

The neural stage and its models

Three controls shape what the Deepfilter stage does. Atten lim caps how far the model may pull the noise down, in dB — lower sounds more natural, higher sounds quieter, and 100 dB means no limit at all. Post-filter β smooths what is left, in small amounts; too much makes speech dull. Mix (dry/wet) blends the cleaned signal back against the original. The stage reports the signal-to-noise improvement it is achieving beside the meters, so the setting can be judged rather than guessed. If a voice starts to sound as though it is underwater, lower the attenuation limit rather than reaching for the mix.

Two DeepFilterNet3 models ship, and every instance picks its own.

The low-latency model is not the compromise its name suggests: it removes more of every kind of noise and takes less of the wanted speech with it. What it costs is processing time. Because each instance runs its own copy of the model, eight of them cost roughly eight times one — about 70 % of a core on the standard model against about 178 % on the low-latency one. Mixing them is normal: the fast model on a telephone hybrid, the full one on a studio microphone. Setting an instance to No model turns the neural stage into a pass-through and leaves the rest of the chain working.

The workspace

Each instance gets a tab, labelled with the card it is capturing from, and every instance keeps processing whichever tab is in front — switching is a display matter, never an audio one.

Inside a tab there are three columns: the meters on the left, the nine stages with their live values in the middle, and the parameters of whichever stage you selected on the right. Click a stage to open it and click it again to close it. The divider between the last two columns can be dragged, double-clicked to reset, or nudged with the arrow keys.

Seventeen colour themes, light and dark, and the interface language are stored on the server rather than in the browser — along with that divider position — so the same view follows you to whichever workstation you open it from.

Metering

Every instance carries two meters, IN ahead of the chain and OUT after it, switchable between peak-programme and loudness so the same pair answers either question. Above the stage parameters both sides are read out together: peak in dBFS, integrated loudness in LUFS, and the signal-to-noise improvement the neural stage is delivering at that moment.

All of it updates about twenty times a second, pushed from the audio engine as it runs rather than polled by the browser.

Running it unattended

Deepfilter is a background service. It starts with the machine, claims its cards and keeps going whether or not a browser is open anywhere. Changing the number of instances restarts every stream, so it is a decision to make before going on air rather than during; changing one instance's card or model restarts only that instance.

Access follows three rules. A browser on the machine itself is always let in, with or without a password, so you cannot lock yourself out. From anywhere else a password is required — and until somebody sets one on the machine itself, access from the network is refused outright rather than left open. Changing or clearing the password signs every logged-in browser out at once.

The interface looks for a new version when you open it and when you ask it to, and at no other time. There is no registration and no usage tracking: if nobody opens the interface, nothing leaves the machine.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided "as is," without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Transcript turns speech into text — from a file you already have, or live from your sound card — entirely on the computer in front of you. Recognition runs against a model stored on the machine, so it keeps working with no internet connection and your recordings are never uploaded anywhere. What comes back is an editable transcript in timestamped segments with the speakers already separated, ready to correct, summarise and export.

Two ways in

Import an existing recording with the button at the top of the files list, or drop one or several onto the list from Explorer or Finder. Audio and video both work — a video file is transcribed from its audio track — and so are the .dct files a professional dictation device produces, decoded internally with no separate conversion step.

Importing deliberately does not start anything. A new file arrives as Queued, and clicking it only selects it, loading it into the player and the waveform so you can listen first. Transcription starts when you press the button for it, which means adding a long recording costs nothing until you actually want the text. Only one file is transcribed at a time; start a second and it waits its turn rather than competing for the same processor and graphics card.

Recording live

Press Record and speak. While the take runs, the files list is replaced by live metering — a goniometer for the stereo image, and a meter with peak-programme bars and a phase-correlation strip on one tab, loudness with current and maximum readouts on the other. Pause and resume as you like; paused time is not written to the recording.

The text appears a few seconds behind you as you speak. Do not expect word-by-word captioning: audio is transcribed in short windows split at natural pauses, so it arrives roughly sentence by sentence. Press Stop and the take is saved as a new file with that text already attached as its transcript — no second pass over the whole recording. Because each window is recognised without the context of the ones before it, live text can be a little less consistent than a finished file transcribed in one go; if a particular recording matters, select it afterwards and run it again with full context.

The transcript

A transcript is a connected list of segments, each with its own timestamp and the speaker who said it. Click into the text and correct whatever the model got wrong, exactly as in any text field. Corrections save themselves a moment after you stop typing — there is no Save button, and nothing is lost by switching files or closing the app.

Two details make correcting fast. Segments already played are drawn in the accent colour, matching the played part of the waveform, and the segment playing right now is highlighted, so the text follows the audio by itself. And clicking a segment's dot or timestamp jumps playback to that moment, while clicking its text deliberately does not — that is how you place the cursor, so correcting a word never moves the audio out from under you.

Speakers

When a transcription finishes, the audio is also analysed for who was speaking. Segments are grouped by voice and each speaker gets a coloured chip beside the text. Click a chip to give it a real name, either for that one segment or for every segment currently attributed to that speaker.

A toggle in the Text bar hides the chips again, and it does more than change the display: with speakers hidden they are also left out of every export and of the summary, so a single-voice dictation comes out as clean prose while an interview keeps its labels. Detection is automatic and approximate — it works from pitch and tonal character, so it can merge two similar voices or split one that changes markedly. Treat it as a starting point to correct, not a finished attribution.

Language, model and your own vocabulary

Pick the spoken language at the top of the files list — English, German, Spanish, French, Italian, Dutch or Portuguese — or leave it on Auto and let the model decide from the first moments of audio. Auto is reliable on longer, clean recordings and can guess wrong on short or noisy ones, and a wrong guess produces nonsense however good the audio is, so if a result looks like gibberish the language picker is the first thing to check.

The recognition model is chosen in configuration, from a catalogue the app downloads for you — no browser, no manual copying. They run from a 75 MB model for quick tests up to a 3.1 GB one for the highest accuracy, and the middle of that range is where most work happens. A larger model is better on accents, background noise and technical vocabulary, and noticeably slower.

The cheapest accuracy win is the glossary: a comma-separated list of names, companies, products and specialist terms, handed to the recogniser as context so it spells them the way you intend. Those are exactly the words a general model gets wrong, and a dozen of them cost nothing in speed.

Export, and an optional summary

Export writes exactly what is on screen — your corrections and your speaker names, not the model's original output. A native save dialog offers five formats: HTML for a formatted document, plain text for further processing, RTF for Word, LibreOffice or Pages, and SubRip or WebVTT subtitles with proper timings for a video editor or a web player. With the speaker toggle off, subtitles come out plain.

A summary is optional and separate. It sends the transcript as currently shown to an Ollama server you run yourself and configure explicitly, and shows the short summary it writes back above the transcript, in whichever language the interface is set to. No such server ships with the app and none is contacted unless you set one up. It is worth being clear about this: summarising is the one feature that sends transcript text off the machine. Everything else — recognition, editing, speakers, export — happens locally, and without a configured server the rest of the app is entirely unaffected.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Player plays broadcast video and audio files and puts the meters a control room needs right beside the picture: a peak programme meter with phase correlation, a loudness meter to ITU-R BS.1770, and a goniometer. Open a file, watch it, listen to it and measure it in the same window — whether that is a clip checked before it goes on air, a delivery measured against its loudness target, or a soundtrack whose stereo image looks wrong.

Opening a file

There are four ways in: the Open Media File… button, dragging a file onto the window, the Recent Media Files list, or the command line. The file dialog lists the common video formats — MP4, MKV, AVI, MOV, WMV, WebM, FLV, TS, M2TS, M4V, MPG, MPEG and VOB — and audio in MP3, WAV, FLAC, AAC, MP2, Ogg, Opus and M4A. A file with any other extension can still be dropped onto the window or passed on the command line, and plays if its format can be decoded.

The player holds one file at a time: opening a file replaces the one that is open. Recent Media Files keeps the last twenty files, the most recent at the top, and one click there opens a file and starts it. Whether a file from the dialog or from drag and drop plays at once is up to you: Autoplay on Open in the configuration decides whether it starts immediately or waits for Play.

When the player is started from the command line with a file, it opens the file and plays it at once; the address of a network stream works there too. If the player is already running, a second start opens no second window: the file goes to the running player, which comes to the front and plays it.

Playback

The transport bar carries Play / Pause, Stop, jumps of ten seconds back and forward, and Step 1 Frame, which shows the next frame while playback is paused. Playback Speed runs from 0.25× to 2×. When a file carries several audio tracks, Audio Tracks picks the one you hear and the meters measure, and Subtitle Tracks appears when a file brings subtitles. With Loop switched on, a file starts again from the beginning each time it ends; otherwise playback stops at the end, Play starts the file again from the beginning, and moving the seek bar starts it from the point you choose.

Click anywhere on the seek bar to jump there, or drag along it; holding the pointer over the bar shows the time at that point. The main controls have keys as well: Space plays and pauses, the left and right arrow keys jump ten seconds back or forward, the up and down arrow keys change the volume, M mutes, . steps one frame, S stops and F switches full screen.

The volume control sets only what you hear. The meters measure the programme ahead of the volume control, so turning the listening level down, or muting it altogether, leaves every reading unchanged.

Picture, aspect ratio and full screen

Video plays with hardware acceleration, scaled to fit the picture area. When a file declares the wrong shape — a 16:9 picture squeezed into 4:3, for example — the Aspect Ratio badge puts it right: besides Auto, which uses the shape the file declares, it offers 16:9, 4:3, 21:9 and 1:1. A choice made with the badge applies to the file that is open; every newly opened file starts with the default set in the configuration.

F or the Fullscreen button shows the picture alone on the whole screen, with every bar and the meter column out of the way. Esc or F brings the window back, and the keys keep working in full screen too. An audio file has no picture, so the picture area shows a turntable with the file name instead: a click there pauses or resumes, and a double-click switches full screen.

Peak level, phase and the stereo picture

The meter column runs down the right-hand side of the window: the goniometer at the top, and below it either the PPM or the LUFS meter, chosen with the two buttons at the top of the column; the player remembers the choice. The meters measure the programme as it plays, in step with the sound. A multichannel soundtrack, 5.1 for example, is measured as its stereo downmix.

The PPM shows the sample peak of each channel in dBFS, on a scale from +10 down to −60: green below −18 dBFS, yellow from −18 to −3 dBFS and red above −3 dBFS. The bars rise instantly and fall at 11.7 dB per second, and a peak marker holds the highest value for 1.5 seconds. The strip beneath is the phase correlation, from −1 to +1: +1 means both channels carry the same signal, 0 means unrelated channels, and values towards −1 mean the channels work against each other, so the programme loses level or cancels out when it is summed to mono.

The goniometer draws the two channels against each other: what they have in common on the vertical axis, what differs between them on the horizontal one. A vertical line is mono, a line along one of the diagonals is sound on one channel only, and a cloud taller than it is wide is ordinary stereo. A horizontal line, or a cloud wider than it is tall, means the channels are out of phase.

Loudness to ITU-R BS.1770

In the LUFS view the two bars show each channel's sample peak from −50 to 0 dBFS, and the readings beneath them show the loudness. Momentary is the loudness of the last 400 milliseconds, steadied over about one second so that it can be read. Integrated is the loudness of everything measured so far, gated as the standard prescribes, with an absolute gate at −70 LUFS and a relative gate at −10 LU. EBU R128 sets −23 LUFS as the target for broadcast, which makes the integrated reading the one to hold a delivery against.

The time beneath the readings shows how long the current measurement has been running. Reset starts a new measurement, and so does opening a file. To measure a single passage, go to where it begins, press Reset and play to where it ends: the integrated reading is then the loudness of exactly that stretch.

Audio output and configuration

The configuration opens from the menu at the top right, and every change takes effect immediately — there is no Save button. For playback it holds Autoplay on Open, the default Aspect Ratio and the Default Meter Mode.

For the audio output there are two systems. WASAPI, the standard Windows audio system, reaches every output Windows lists. ASIO talks directly to a professional audio interface through the interface's own driver, and two more settings then appear: the pair of the interface's outputs that carries the sound, and the sample rate. 48 kHz is the broadcast standard and the rate the player works at, so at any other rate the sound is converted once on its way to the interface. The player falls back to the default output if the saved one is missing at start, and to WASAPI if no ASIO device can be found.

Updating or uninstalling the player leaves the settings in place. The player checks for a newer version each time it starts, and a click installs it. Seventeen colour themes, light and dark mode, and seven interface languages are built in.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Reporter is the workstation for the person who writes the piece, cuts the tape and files it. Four tools that usually mean four programs share one window: a script editor that tells you how long the text runs, the Audioeditor for the cuts, a broadcast phone line for the interview, and a mixer for everything the machine can hear. Coming soon.

One window, four tools

The studio is a docking grid, not a stack of windows. The Script Editor sits top left, the Audio Editor top right, and the Mixer runs across the bottom — all three live at once, so a cue you write is a cut you can hear and a fader you can pull without leaving the screen.

Drag any tab to an edge to split the view, or into the middle of another panel to stack them as tabs. A panel can be maximised to fill the studio and restored again with the same button. Whatever you end up with is remembered and comes back the next time the application starts; one button in the header puts the standard three-way layout back.

The script, with its clock

The editor is built for reading aloud rather than for print. As you type, the bar under the headline counts words and characters and turns them into an estimated read time at 130 words per minute — the international radio reading speed — shown as minutes and seconds. A bulletin that has to end on the hour can be written to fit rather than trimmed on air.

Quotes and soundbites belong in the text, so they go there: Insert audio cue drops a marked pill into the manuscript carrying the clip's name and its length, which is where a presenter needs it when the fader has to go up. Scripts are kept as individual documents and the Story Archive slides out to bring an older one back.

Cutting the audio

The audio panel is the Audioeditor itself, filling it rather than sitting in a window of its own. Zoom into the wave, scrub the timeline, cut, copy, paste, trim and take out the silences.

Beyond editing there is level and repair work: gain normalisation to the EBU R128 loudness standard, fades in and out, phase inversion, reverse, and a parametric equaliser. Drop a WAV, MP3, OGG or FLAC file straight onto the wave to load it, and export the finished piece as broadcast WAV at 48 kHz in 16 or 24 bit, or as MP3.

The line to the studio

A report from the field needs a line home, so Reporter carries a broadcast codec rather than a consumer voice call: a native SIP engine following EBU Tech 3326, the N/ACIP specification broadcasters use to connect one studio to another.

The call can be recorded while it runs, uncompressed, and the recording opens in the Audioeditor with one click — the interview is ready to cut before the guest has hung up. The dialer slides in from the edge with a keypad, several registered lines to choose between, and a searchable phonebook.

Contribution over the network

Where a phone line is not enough, Reporter is also a full contribution client for audio and video. Remote contributors, field reporters and studio guests appear on a video stage that grows to fit them, with whoever is speaking marked; the audio is carried with echo cancellation, noise suppression and automatic gain so an untrained guest on an ordinary computer still sounds usable.

Camera and screen sharing let a rundown or a presentation be shown to everyone in the room. A roster lists who is connected, whether their microphone is open and how good their connection is, and a chat channel carries editorial cues without interrupting the audio. The whole conversation can be recorded to a broadcast file in one click and edited immediately.

The mixer

Along the bottom is a real console rather than a volume slider. Every audio input the computer can see gets its own strip, whatever it is — a USB interface, a virtual sound card, an analogue microphone input — and the phone line, the contribution feed and the master bus get their own alongside them.

Each strip is built the way a broadcast desk is: PFL to hear a source in the headphones without putting it on air, a fader with unity at three quarters of its travel and a scale from −60 dB to +10 dB, a double-click on the cap to drop straight back to 0 dB, peak meters to EBU R68 beside the fader with a shared zero line, and a large ON button at the foot that is unmistakable across the room. The master carries a soft-knee limiter so a sudden laugh cannot clip the file.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided "as is," without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

A studio is built from equipment by a dozen manufacturers, and almost none of it speaks the same language. Digimedia Smarttool sits in the middle and translates: it connects to consoles, playout systems, video switchers, signage players, lighting and sensors, maps what each one exposes into a single tree of parameters, and lets you wire any of them to any other with a rule. It runs as a background service and is operated from a browser. Beta phase running.

Five words

Everything in Smarttool is built from five ideas, and knowing them is most of learning it.

Add a device, open its tree, and tick Rule on a parameter to make it a source or Store to keep its value. Subscriptions are handled for you: ticking a box subscribes on the device, clearing it unsubscribes again once nothing else needs it.

What it can talk to

Each device type is its own connection, with the right defaults filled in — you supply an address and, where the protocol wants one, a credential.

Anything speaking a bidirectional message protocol that is not on this list can still be added generically, with its messages taken apart into a tree automatically.

Rules

A rule watches one parameter and writes to one or more others whenever it changes. Source and destination need nothing in common — a fader on a console can move a register on an industrial controller, and one rule can drive several destinations at once, each with its own handling.

That is what makes one gesture do several things: opening a microphone fader can light the red sign, cut the camera to that presenter and dim the studio monitor, from a single rule. Each rule has a play button that runs it against a value you type, so the logic can be proved without touching the hardware.

Variables and the clock

Parameters are momentary; variables remember. A stored value stays in the local database across a service restart and a reboot, and any rule script can read or write it — which is how one system's state becomes another's, and how a rule knows what happened before it ran. A variable still referenced by a rule cannot be deleted by accident; Smarttool names the rule that is using it.

The scheduler fires a rule on the clock rather than on an event: at an exact date and time, on a repeat from minutes to months, or on a plain cron expression, with the visual editor and the expression kept in step. That covers the station's morning start-up, the nightly reset, a periodic transmitter check.

Logs

Rules and tasks can write an entry every time they run, which turns "the light did not come on" into a line with a timestamp. Entries are graded by severity, carry local time and can be filtered as you type.

The same view is also a collector for the rest of the system: point network switches, audio routers, playout servers and Linux hosts at Smarttool and their messages land in the same searchable place, instead of in a dozen separate log viewers nobody opens.

Running it unattended

Smarttool is a Windows service. It starts with the machine, holds its connections and keeps running with no one logged in and no window open; the browser interface is only a remote control, and closing it changes nothing. Values arrive as they happen rather than being polled, so a tally is live rather than a second old.

Access is closed until you open it. A browser on the machine itself always gets in, so you cannot lock yourself out. From anywhere else a password is required — and until somebody sets one on the machine itself, access from the network is refused outright rather than left open, which means a freshly installed service is never exposed by default. Changing or clearing the password signs every remote session out at once.

The service looks for a new version shortly after it starts and then only when you ask it to, so it is not talking to the outside world during a show.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided "as is," without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Visualradio turns a radio show into a watchable one without putting anybody behind a vision mixer. It listens to each microphone, watches the fader and on-air state of every channel on the console, works out who is actually talking, and cuts the cameras accordingly — presenter, guest, wide shot — driving whatever switcher or PTZ camera the studio already has. It runs as a background service and is operated from a browser. Beta phase running.

What it listens to

Two things decide who is on camera, and both have to be true. The first is sound: every microphone channel has its own gate, opening at around −34 dBFS by default, so breathing and headphone spill never count as speech. The second is the console: a channel is only a candidate while its fader is actually open and the channel is on air. A presenter leaning into a closed microphone changes nothing.

Fader moves, on-air states and level readings arrive as they happen rather than being asked for at intervals. That is what makes a cut land on the word rather than a beat after it.

The rules that stop it twitching

An automatic director that simply followed the loudest microphone would be unwatchable. Four rules turn level readings into something that looks directed.

The pace is adjustable while you are on air: hold the shots longer for a relaxed talk segment, shorten them for a fast morning show.

Which microphone is which camera

One table connects the studio to the video. Each row ties an audio channel and its fader on the console to a camera: the input on the switcher, the scene name if the switcher uses scenes, and for a PTZ camera the preset to recall, so the camera points at the right seat before it goes on air.

Rows also carry a priority, which settles the case nobody plans for — two people starting to talk in the same instant. Without it the director would pick arbitrarily; with it the outcome is the one you decided in advance.

What it can drive

The director does not replace the studio's video equipment; it operates it. Whatever is already there is what it cuts between.

The control view

The operator's screen is two panes side by side: what is on air now, in red, with the shot, how long it has been on air and how long it is held for; and what the director intends to take next, in green. Watching the second one for a minute is how you learn to trust the first.

None of this has to stay automatic. Take any camera by hand at any moment, freeze the picture where it is, or hand control back to the director — and the pace can be changed on the fly without stopping. Manual takes are always available, so an operator who wants a particular shot simply takes it.

In the studio — and left alone

A second screen is meant for the people in front of the cameras: a large tally for a confidence monitor or a tablet, red for whoever is on air, amber when the wide shot is up, a timer, and a cue telling the next presenter that their camera is coming. That screen answers the first question anyone in a radio studio asks, which is whether they can be seen.

The service starts with the machine and runs unattended. A browser on the machine itself is always let in; from anywhere else on the network access is refused outright until a password has been set locally, so it is never quietly left open.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Scheduler decides what goes on air, hour after hour, days and weeks ahead. You describe the station's format once — which categories of music exist, how often each may come back, what a typical hour is built from — and it assembles the broadcast day from that, one song at a time, and keeps every separation rule you set. The finished day goes straight to Digimedia Playout. It runs as a background service and is operated from a browser. Coming early 2027.

What the scheduler knows about a song

The library holds the music, the jingles and the promos, and every entry carries the handful of facts the engine actually reasons about. Fill these in once and every decision below follows from them.

Categories, and how often a song comes back

Categories are rotation pools, and the split between them is what a listener hears as the station's format. A typical set runs from the current power hits through recurrents and two tiers of gold, alongside pools for station imaging and for commercial breaks.

How often a song returns is not a setting — it falls out of arithmetic. Divide the number of active songs in a category by the number of positions per hour your clocks give it, and you have the number of hours before a song comes round again. Forty songs in a pool that gets two positions an hour turn over in twenty hours.

Aim for an odd number of hours. A category that comes round every six or twelve hours lands on the same songs at the same time of day, and the listener driving to work at eight hears the same record every morning of the week. At five, seven or eleven hours it drifts across the day instead.

The rules that keep the hour listenable

Rules are what stop a correct-looking playlist sounding wrong. Two of them are absolute — the engine leaves a hole rather than break either. The rest carry a penalty instead: the engine avoids breaking them where it can, and breaks the least costly one where it must.

The hour as a wheel

An hour clock is the blueprint for sixty minutes: not the songs themselves, but the shape they have to fill. You lay the positions out in order and the editor shows the same hour beside them as a dial, every element a coloured segment, with the running total against the sixty minutes — so you can see at a glance whether the hour is short or over.

A position is one of a small set of kinds: music drawn from a category, a jingle, an ident, a sweeper, a bed, a commercial break, news, a voice track, or something of your own. A music position names a category rather than a song, which is why the same clock produces a different hour every time it runs.

Some things have to happen at a set moment — the news at the top of the hour, a network join, a fixed feature. Pin a position to a time and the engine works backwards from it and chooses what comes before, so that the hour arrives there exactly rather than approximately.

A day, then a month

Twenty-four clocks make a day template. Most stations need only a handful — a morning drive hour, a workday hour, an afternoon drive hour, an evening hour and something to run overnight — and the template says which clock each hour of the day gets. Weekdays, Saturdays and holidays become separate templates without duplicating the work.

The calendar puts the month in front of you and colours every date by how far along it is: nothing assigned yet, a template assigned, the day generated and waiting to be looked at, or published and gone to playout. It is the one screen that answers whether next week is ready without your having to open anything.

Building the day, and handing it over

Generating a day fills every position in every hour. For each one the engine takes the category's candidates, discards everything that would break an absolute rule, scores what is left against the rules that carry a penalty and takes the best-scoring candidate — then updates how long ago that artist and that title were heard, so the next choice knows about this one. Twenty-four hours resolve in seconds.

Nothing is hidden afterwards. The plan marks every position where a rule had to be relaxed, and clicking a song offers the alternatives ranked by how well they fit, so a substitution takes a moment and you can see what it costs. The analysis view answers the wider questions: how the hours divide between music, imaging and commercials, which artists get too much or too little airtime over a run of days, and where the engine had to give way to avoid an unfillable hour.

Publishing writes the finished day into the folder Digimedia Playout watches, and it is picked up from there. Everything the buttons do can also be driven from another system, so a nightly job can build tomorrow and publish it without anyone opening the browser.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements

Digimedia Producer is where content for podcast and broadcast is planned, created and published. The work starts in the newsroom: feeds, news-agency copy and email bring in the ideas, and an AI server you run yourself helps weigh them and draft the presenter text. From there, content is organised on five levels — the episode, the segments it is built from, the stories they tell, the social posts derived from them, and the individual media items underneath. All of it moves across a production board and onto a broadcast calendar, is cut in studios built into the app itself, and goes out to Apple Podcasts, Spotify and other podcast directories, to YouTube and to your playout system. Coming Q1 2027.

Content for podcast and broadcast

What a station publishes does not live on a single level, so Producer does not work on one either. The same production is described and edited at five tiers, each with its own fields and its own place in the workflow.

Newsroom research

The Research tab puts the day's material in the same window as the production. A source list on the left groups your feeds by kind and lets you switch each one on or off; the stream in the middle can be searched and narrowed to what is still unread; the reader on the right shows the article, its author and a link back to the original. RSS and Atom feeds are supported, and the list of sources is yours to extend.

Taking a story over is one click. It becomes a card for the channel you picked, in the idea-and-research column, with its source attached: where the story came from, when it was published, the link and an excerpt of the original. A second button does the same and adds the story to the script as a presenter segment. Weeks later, anyone who opens the card can still see where the idea began.

Ideas also arrive without anyone opening Research. Point Producer at a mailbox and every new unread message becomes an idea card, titled with its subject line. NewsML files from Keystone-SDA, dropped into a watched folder, do the same, with any pictures, audio or video that come with them attached to the card.

An AI that knows your station

Producer comes with no AI service of its own. Instead it talks to a server that you run yourself — Ollama, or a vLLM server. You enter its address once under Configuration, where a test shows whether it answers and how quickly. The AI is entirely optional: without a server, the rest of Producer is unaffected.

What makes the answers fit each channel is Format Memory. Each channel keeps its own station profile — the tone of voice, the audience, the longest a spoken intro may run and a list of style rules — so a hit-radio channel and a news channel can sit side by side and get very different copy, each written to the rules of the channel the story is for.

Three jobs are built in. On an incoming story, the AI scores the story's relevance for the channel from one to ten, says why in one sentence, picks out the key facts for the presenter and suggests an interview question. It writes the presenter's text — an intro or a short teaser — to be heard, not read, in the channel's tone and within the channel's time limit, with the reading time worked out at about 130 words a minute; the same text generator sits inside the script editor, one click from a segment. And the AI drafts the social side: a post with a hook, a question for the community and hashtags, and a script for a vertical video with hook, story and call to action.

It only ever suggests. The result is text you can edit, copy or drop into the script, and the AI never puts anything on air or out to a feed itself.

The board and the calendar

The board is the editorial pipeline, laid out as five columns that a piece of content moves through: idea and research, to produce, in production, ready to air, published. Every card shows its channel, its air date, how much of its checklist is done and who is on it, so the state of the whole week is one glance rather than a round of questions.

The calendar is where those cards become air dates. Everything without a date waits in a pool down the left-hand side; drag one onto a day and it is scheduled. Days list their shows in order with channel colours and status badges, and clicking a badge opens that card's workspace directly.

The content workspace

Opening a card gives a flat two-column workspace instead of a stack of dialogs. On the left: the title, the outline, an interactive production checklist and the media items attached to it, each with its own waveform player. On the right: the status, a finished toggle, the air date with quick presets for today, tomorrow and next week, the channel, the host with their own colour, a completion gauge, and shortcuts straight into the studios. A card that started in Research shows its source at the top, with a link to the original.

A drawer behind the hamburger at the top left searches every card and lists them with their live completion gauges, so moving from one piece of content to the next never means going back to the board first.

Script, rundown and teleprompter

The script editor breaks a show into segments — intro, host talk, audio take, jingle, interview, outro — and adds their durations up as you write, planned duration against actual. Reading time is calculated from a words-per-minute rate, so a script that overruns is obvious while there is still time to cut it rather than on air. Each segment names its speaker and carries its own cues.

Segments reorder by dragging, either as a vertical list or along a horizontal timeline. When it is time to present, a full-screen teleprompter shows the script in high contrast, and a print view lays the same rundown out for presenters who would rather hold paper. Templates give a new script a running start — there are ready-made ones for a 30-minute show, a 60-minute magazine and a 45-minute interview podcast.

The audio and video studios

Click an audio item and it opens in the Audioeditor: parametric and graphic EQ, compressor, reverb, delay and distortion, and EBU R128 normalisation to −14 LUFS for streaming or −23 LUFS for broadcast. A microphone records a voiceover straight into the open project, inserted or appended, and saving the take writes the processed audio back onto the card it came from.

Video gets a multi-track timeline: titles, main video, B-roll, voiceover and music, with broadcast lower thirds and subtitles over the top and a master render out to 1080p or 4K MP4. Neither studio is a separate application to launch and save out of — both are part of the card you already have open.

Transcripts from your recordings

Drop a recording onto a card and Producer gets to work on it in the background: thumbnails for video, and a speech-to-text transcript that streams in with timecodes as it is produced rather than arriving all at once at the end. The interface stays responsive throughout, so you can keep writing the script while an interview is still being transcribed.

The transcript is not only there to be read. It becomes the transcript entry in the podcast feed, and its timecodes are what chapters and soundbites are built from.

Publishing and playout export

Producer writes a Podcasting 2.0 feed for Apple Podcasts, Spotify and every other directory: chapters, transcripts, soundbites and per-person credits, so a modern player shows the episode properly instead of a title and a play button. A new release is announced the moment it is ready over Podping and WebSub, rather than waiting for directories to come and look.

Video is uploaded to YouTube with the chapter timestamps written into the description, and the short-form posts go out on their own schedule. What is going on air stays in house: shows, takes and metadata export to Digimedia Playout and Digimedia Cartwall with their start times, cue points and channel IDs intact.

Configuration and your data

The board is kept as plain JSON on your own machine and synchronised in real time, so a second screen in the studio follows the newsroom without anybody refreshing anything. The overflow menu exports the whole board to a file, imports one back, and resets to demo data when you want to try something without consequences.

That second screen can be filtered to one channel or one presenter, full screen and without the navigation — a ready-made display for the studio wall, or a panel in Digimedia Dashboard.

Access from the network needs a password. A browser on the machine itself always gets in without one, so nobody can be locked out of the box in front of them. Until a password has been set on the machine itself, access from the network is refused outright rather than left open.

The interface comes in seventeen colour themes with a light and a dark mode, and in seven languages. Credentials for publishing — YouTube, cloud storage and podcast directory notifications — are entered once in configuration and stay there, as do the address of the AI server and the list of news feeds.

Licensing

This is a free tool. No obligations to buy anything. Downloads are provided “as is,” without warranty of any kind. Micromedia AG assumes no liability for any damages arising from their use. A licensed enterprise edition with support is available on request. Please contact us for a quote.

System requirements