TG Studios

TG Drop Deck

Eighty sounds a keypress away, a running order that crosses itself, your microphone properly processed, and the whole show out to your own server, to YouTube, and onto the disk. Without touching a mouse.

This is an open beta. It works, and it is what I run my own show on, but the keys, the wording and what it does are still moving while people test it and write in. Once the final version is set in stone, it becomes just Drop Deck.

Free. Version 3.8.2 for Windows and 3.8.0 for Mac. Forty sounds included, so it makes a noise the first time you open it.

No sign-up, no key. The installer puts it in your own user folder, so it never asks for an administrator. If you would rather not install anything, take the zip, unzip it and run TG Drop Deck.exe with the demo folder beside it. A portable copy updates itself in place.

On a Mac, unzip and move TG Drop Deck into Applications. It needs macOS 14 or later on Apple silicon. It is signed and notarised, so it opens like any other app, and it updates itself from inside the app. It is the same program with the same board file, written natively for VoiceOver, and the Mac manual has its keys: the Windows keys with Command where Windows has Ctrl.

Free, and staying free. If it earns a place in your show, there is a donate page, pay what you like, and nothing is held back from anyone who does not.

What it actually does

It started as a soundboard. It runs a show now.

All of it by keyboard, all of it spoken, and none of it behind a screen a screen reader cannot read.

The whole app is the number row

Press 1 and you hear applause. That is the entire learning curve.

Effects and drops overlap and never cut each other off, so a laugh can land on top of a sting. Beds toggle, and only one plays at a time.

Three faders, because one is not enough

Sound effects, music beds and the running order each have their own level, and each says the new one as you move it. F3 and F4 for sounds, F5 and F6 for beds, F7 and F8 for the playlist.

While you are on air, F7 and F8 become a monitor fader: pull the music right down to hear your screen reader and navigate, and your listeners still get it at full level.

The beds get out of the way

Fire an effect or a drop and the music underneath drops about nine decibels, then slides back up when it finishes. It is the thing radio does that makes a show sound produced rather than assembled, and here it happens without you touching anything. Opening the microphone does it too.

Ctrl+D turns it off. Ctrl+P sets how far it ducks.

A running order that runs itself

Ctrl+Shift+P is the playlist. Paste songs in from File Explorer with Ctrl+V, a folder at a time if you like.

Each song hands over to the next before it ends, at a crossfade you set, and any single track can have its own. Untick one and it is skipped without losing its place. Shift+Enter crosses into the track you are on, which is how you get out of a song early, and Alt+Home and Alt+End move a track to the top or the bottom of the order.

It reads artist and title out of the file, shows when each track starts, warns you before one ends with a sound you choose, and drops an ident in from your library every so many songs if you want one. Save the whole order as an M3U and open it again next week.

Your voice, done properly

Ctrl+Shift+M is the microphone: which one, how much gain, which ear you hear yourself in, and whether you hear yourself at all.

The Voice tab is a gate, a high pass filter, a three band equaliser, a compressor and a limiter that means it. Every setting is a row in a list: up and down to choose, left and right to change, and it says the new value as you go. Nothing here opens a window you cannot read.

Load a VST3 vocal effect and its own knobs appear in the same list, in words, with real units, read out as you arrow through them. A plugin whose window no screen reader can touch becomes a list any screen reader can. Presets save and open, and the plugin you chose comes back next time.

Out to your own server

Ctrl+B goes live. Icecast, Liquidsoap or SHOUTcast, in MP3, AAC or Ogg Opus, at the bitrate you choose, with a Test the connection button so you find out a password is wrong before the show rather than during it.

Save as many stations as you run and switch between them from the On air menu. The track title goes to the server as it changes. Ctrl+Shift+B says what the stream is doing, and Ctrl+Shift+A opens a window showing how many people are listening and what is playing, which works whether or not you are on air.

If the connection starts falling behind, it tells you, once, rather than quietly losing audio at the far end.

Video, for the platforms that insist on a picture

YouTube, Facebook and Restream will not take a stream with no video track, so Drop Deck always sends one, and you choose what it is. Alt+Shift+V offers a card drawn from your station name and what is playing, your own artwork, a camera, your screen, or your screen with the camera small in a corner you pick. You can change between them while you are on air, without dropping the connection or breaking the sound.

Alt+Shift+T puts things on top of the picture: your station name, what is playing, a clock, or your own words, in four named places that cannot overlap. Ctrl+Shift+V reads out what is currently on screen, so you always know what a viewer is looking at. If the picture goes black or freezes, it tells you.

Alt+Shift+C sets your colours. Not with a colour wheel, which is no use to anybody who cannot see it, but by name, with every colour telling you how well it reads as you arrow onto it: "gold, easy to read, 8.6 to 1". Ten ready-made looks are offered first and every one has been checked against the numbers rather than chosen by eye.

It can describe your own shot back to you

Alt+Shift+D sends one still of the picture going out to Claude, ChatGPT or Gemini, on your own account, and reads back what somebody who could see it would have noticed: whether you are centred, whether the light behind you is brighter than your face, whether your shirt blends into the wall, whether the words on screen are sitting across your chin. On a screen share it looks for anything private, and says where it is.

You can ask it follow-up questions, and it remembers the conversation.

It is never needed to go live and going live never waits for it. It never sends a picture of your screen without asking, and it asks every single time.

What is new

3.8.2 for Windows gives each destination its own key. Ctrl+B goes live on your video platform and Alt+Shift+B goes live on your radio station, so the summary you are shown before going live is always for the place you meant. It also stops naming your YouTube channel after your radio station: there is a Channel name box of its own on the Video streaming page, and left empty Drop Deck just says the platform.

3.8.1 for Windows fixes what a video recording actually records. If your Ctrl+B was pointed at a radio station, Ctrl+Shift+R recorded a card with your station name on it instead of your camera and your screen, and it would not let you choose the picture at all. Your picture belongs to the app now rather than to a video platform, so Alt+Shift+V, Alt+Shift+T and Ctrl+Shift+V work whether you go out on Icecast, on YouTube or nowhere. Changing the picture reaches a recording that is already running, words on the picture show up on a card, which they never have, and Ctrl+Shift+W reads out your picture and what is using it.

3.8.0 for Windows makes the cable something you can check rather than something you hope about. Test the cable, on Alt+Shift+O, plays a tone down your virtual cable, records the other end of it, and tells you whether it arrived and at what level. Drop Deck also names the exact device to choose in TeamTalk or Zoom, which everybody gets backwards the first time because the two ends of a cable are named from the cable's point of view rather than yours. And it says so when Windows has the two ends set to different sample rates, which is easy to do by accident, costs two conversions on every sample and nothing in Windows ever mentions.

Drop Deck Audio is coming to Windows. The Mac has it now. On a Mac a virtual audio device is a plug in that lives in a folder; on Windows it is a kernel driver, and a kernel driver has to be signed by Microsoft before Windows will load it at all. That is a longer road than one release. Until then Windows uses VB-CABLE, which is free and takes two minutes, and everything above is about making that painless.

3.8.0 for Mac catches the Mac up with everything Windows gained between 3.6.0 and 3.7.1, and adds one thing Windows has not got: Drop Deck Audio, a virtual audio cable that comes with the app. One menu item installs it, and after that it is in every program's microphone list under that name. Set TeamTalk, Zoom or Skype to take Drop Deck Audio and they get your whole show: pads, beds, running order, your microphone with its processing and every source you are catching. It can leave one source out, so sending to a program you are also capturing does not fill the call with echo. Installing it asks for your password and restarts Core Audio, and the app says so before it asks.

The Mac also gets the send itself on Option+Shift+O for any cable you already have, recording the picture on Command+Shift+R, the cue sheet on Command+Shift+C, the .cue track list beside every recording, holding a source back in milliseconds, and Command+Shift+W, which reads out where every single thing is going in one go.

3.7.1 writes a .cue track list beside every recording, with the same file name, so a mix and its tracklist go to Mixcloud together.

3.7.0 records the picture as well as the sound. Ctrl+Shift+R takes video and audio together into one MP4, on air or off, and it stays in step: measured over three minutes of real recording it drifted by a tenth of a millisecond a minute. There is a cue sheet on Ctrl+Shift+C showing what is coming up from your running order, and audio sources can be held back in milliseconds to line up a capture card. Both recording keys now tell you what is actually in the recording.

3.6.0 and 3.6.1 gave your whole show its own output, so TeamTalk, Zoom, Discord or OBS can take it through a virtual cable: microphone, sources and all, with mix minus so you never feed a program its own audio back. Ctrl+Shift+W reads out where every single thing is going.

3.5.2 puts the camera in whichever corner of your screen you like, lets you ask follow-up questions about your shot, and will tell you what your branding actually looks like to somebody who can see it.

3.5.1 and 3.5.0 added the colours, the things on top of the picture, and the shot check, and fixed a colour fault that had been going out on every stream: the picture was converted with the standard definition formula and sent unlabelled, so every viewer's player had to guess.

Every release, with what changed and why, is on the changelog.

Record it, and put other things on the air

Ctrl+R records the show. The same mix that goes out: every sound card, the running order, and your microphone. WAV, MP3, AAC or Ogg Opus, into Documents as Drop Deck Stream 001 and counting up. It does not need you to be on air, and recording while streaming takes nothing away from either.

Audio sources put anything Windows offers as an input on the air with you. A second microphone for a co-host. A hardware mixer. Or another program entirely: point TeamTalk, Chrome or a game at a virtual audio cable in its own settings, choose the cable here, and it goes out with the rest of the show at a level you set.

And it feeds OBS the other way. Send a bank to a virtual cable in Preferences, add that cable in OBS as an audio input, and Twitch or YouTube get the soundboard along with everything else.

What comes with it

Twenty sound effects and twenty looping music beds, applause, buzzers, an air horn, a rimshot, risers and impacts, and half-minute beds from lo-fi and funk to news urgency, ambient and late-night jazz.

Every one of them was generated with AI, using ElevenLabs. Nothing in the pack is recorded or sampled from a commercial sound library, which is what makes it safe to hand out with a free program. Use them for whatever you make.

The beds are trimmed to an exact loop and crossfaded, so they run under you for an hour without a click, and every effect is levelled so nothing jumps out or disappears.

Then make it yours

Ctrl+N empties all eighty slots. Land on any button, press Space and pick a file: WAV, MP3, M4A, FLAC, OGG, WMA, AAC, Opus and more, so an iTunes or Apple Music library goes straight in. F2 names it so you recognise it live. That is the whole workflow, and the board saves itself.

Turn on preview in the file browser and each sound plays as you arrow onto it, so you pick by ear rather than by file name.

Sounds are never copied into the program; it remembers where they live. Move your sound folder later and File, Relink missing sounds repairs the whole board, the running order and your drops library against the new location in one pass.

Ctrl+S saves a board to its own file, so you can keep one per show and open whichever you need. File, Load the demo pack brings the included sounds back at any time.

Send it wherever you need

Ctrl+P chooses the output device, including a virtual audio cable, and each bank can go to a card of its own: beds to one output, effects to another, the running order to a third. Feed a stream or a recorder while you keep listening on your own speakers.

Devices are remembered by name, so plugging in a headset later will not silently move your sound somewhere unexpected, and if one is missing when you open the app it says so rather than playing out of the wrong card.

Where it came from

This is a rebuild of the soundboard I have used on my own show for years. The original had no surviving source code, so the keyboard map was recovered out of the program itself before a line of the new one was written, which is why anyone who used the old one already knows this one.

Underneath it is new: a real mixer, proper sample rate conversion, long files streamed from disk instead of loaded whole, and ducking that was not there before. Everything since, the playlist, the microphone processing and the streaming, came from people who use it and wrote in.

Requirements

Windows 10 or later. No Python needed. Built and tested with NVDA in mind; if you use JAWS and try it, tell me how it goes.

A Linux build is being talked about. Nothing is promised yet.

Free and MIT licensed, and it stays free and MIT after the beta. The whole source is on GitHub, so nothing here depends on me still being around. Windows may warn the first time because it is not code signed, choose More info, then Run anyway.