Sound from words
Describe a sound. Keep every take.
Pick a voice-over, a music bed, a sound effect or a two-voice dialogue, press Generate, and the takes appear one by one with their waveforms in a history that keeps every run. Make variations, keep collections, upload your own and save to Geenius OS Media with provenance.
Request access What makes the sound
Early access, and plain about it: today the sound is made by a Preview that is simulated in your browser, a robot voice and synthesised beds, not by a voice model. The windows on this site show sample data.
Made bya simulated Preview todayvoice models where a key is set
Before anything else
What is simulated.
We would rather say it here than let you find out in the app.
- Pre-betaThe Preview is the maker today
It synthesises real sound in your browser in plain code: a robot voice built from formants, music beds from chords, bass and drums, and effects shaped by words such as rain, whoosh or boom. It is real sound, and it is not a voice model or a natural human voice.
- Pre-betaReal voice models are connected in code
Gemini speech, ElevenLabs voices and an optional OpenAI voice are wired for single voices and work only where their keys are set on the server. Whether any is switched on for the hosted app is to be confirmed.
- PlannedMusic and effect models
Google Lyria music and ElevenLabs sound effects are planned, not connected. Two-voice dialogue with a real model is planned.
What it does
Four kinds. One history.
The composer, the history, the takes and the library are built and in the app. They work the same with any provider, and each take says which one made it.
Voice-over
English, German or Dutch, with a voice, a speed and an emotion.
Voice-oversMusic and effects
A music bed with a mood, a length and a loop; a sound effect up to 15 seconds.
MusicDialogue
Two voices from a script, one line per turn.
DialogueTakes
One to four takes a run, each with a waveform you can play.
History
The takes
Listen, then keep the one you like.
A run makes one to four takes. They appear one by one, each on its own waveform you can play. Make variations of a take, reuse a run's settings, rename and download in WAV or MP3, or upload sounds of your own.
- Peaks drawn for every take
- Variations with the same settings and new seeds
- A failed run is refunded the takes it never made
Credits
The cost before you press.
A voice-over or a dialogue costs 1 credit per 500 characters, music 1 credit per 10 seconds and an effect 1, times the takes. The composer shows the cost before anything is made. Premium has 1,000 credits a month and Max 3,000; Pro makes no sound.
- Counted per calendar month
- A failed run is refunded
- The Preview costs no Geenius OS credits
Geenius OS Media
Saved with provenance.
Save a sound to your Geenius OS Media library together with its words, model, seed and settings, or bring a sound in from Media. Collections and uploads keep the rest in order.
- The provenance travels with the sound
- Uploads are kept in the library
- Downloads any time
A door for your AI
Your AI queues. A page makes.
Connect Claude, ChatGPT or Grok to Geenius OS once. Your AI can ask which providers are ready, queue a run, read the takes and keep collections. Because the Preview runs in a browser, a queued run is made as soon as somebody in the workspace has AI Audio open.
- Two tools: audio and collection
- On every plan, making on Premium and Max
- It cannot promise a result
In numbers
Facts you can check, not testimonials.
AI Audio is in early access and has no public customers yet. These are things you can look up in the app and the docs today.
- 4kinds: voice-over, music, effect, dialogue
- 3languages for speech: English, German, Dutch
- 4takes at most in one run
- 1provider ready for everyone: the simulated Preview
Honest labels
What works today.
We say what is live in early access, what depends on your setup, and what is not offered.
- Pre-betaComposer, takes, history, library and Media
Built and in the app now. Early access: people request access.
- Pre-betaYour AI through Geenius OS
Works when AI Audio is added to your Geenius OS workspace, on every plan.
- Pre-betaReal voice models
Connected in code and ready only where their keys are set. We do not claim they are on for you.
- Not offeredMaking sound on Pro
Pro is the library. Premium and Max make sound.
- Not offeredA phone app
Web only for now.
Who it is for
For people who want a library and a maker.
Scratch tracks and mock-ups
Simulated voices and beds for timing and drafts, never passed off as real.
See the setupSound libraries
Uploads, collections, search and downloads for sounds and their settings.
See the setupGeenius OS users
Add sounds from Geenius OS Media; save results back with their provenance.
See the setupMakers with their own AI
Your AI asks what is ready, queues a run and keeps collections.
See the setup
Pricing
Three plans. No free tier.
Per month, in euros, VAT included. Included in the matching Geenius OS plan. Payments are in test mode during early access.
- €9Pro: the library, no making
- €19Premium: 1,000 credits a month
- €49Max: 3,000 credits a month
From the blog
Notes on sound, honestly.
A robot voice, and why we say so
What the AI Audio Preview voice is, how it is made, why it sounds like a robot and why that is the honest starting point for sound from words.
Check the cost first
How AI Audio shows the credits a run costs before anything is made, how credits are counted per take, and what is refunded when a run fails.
Take four, keep one
Why AI Audio makes one to four takes per run, how to listen on a waveform, and how to keep the one you like with its settings.
Questions
Good questions.
What is AI Audio?
A library and a maker for sound. You describe a voice-over, a music bed, a sound effect or a two-voice dialogue, and the takes appear with waveforms in a history that keeps every run and its settings. Its full name is Geenius AI Audio. It is an app of Geenius OS and also works on its own.
Is the sound made by an AI model?
Not by default today. The Preview that makes it is simulated in your browser: a robot voice, synthesised beds and word-shaped effects. A real voice model is connected in code and works only where its key is set on the server.
Will the voice sound human?
Not with the Preview: it is a robot voice. A real voice model may, where one is switched on.
Is it ready to use?
It is in early access: people request access and payments are in test mode. The composer, takes, history, library and Media are built. Request access.
Who makes it?
AVS Media, a small company in Nijmegen, the Netherlands, as one of the apps of Geenius OS.
How does my AI use it?
Through Geenius OS: tools called audio and collection. It can check providers, queue a run, read takes and keep collections.
Describe the first one.
AI Audio lets people in step by step. Ask for a place, or sign in with Geenius OS.