Download
Up and running in five steps.
Acousma is a Windows desktop app. Everything below happens on your machine.
Free · No account or sign-up
Platform
Windows 10 / 11 · x64
One-time download
~350 MB speech models
For the assistant
An API key, or a local server
- 1
Get the executable
Download
Acousma.exe— a single self-contained ~75 MB file, no installer and no separate .NET runtime. - 2
Download the speech models
On first launch the app notes the models aren’t installed. Press ⬇ Download models in the toolbar; it fetches ~350 MB once into a
models/folder beside the exe, then the button disappears. - 3
Set your input level
Open the audio menu, pick your microphone, and drag Input gain so the live meter sits below the red clipping zone. Clean input is the single biggest lever on transcription quality.
- 4
Choose your assistant
In Settings, pick a provider. Paste an API key for Claude, OpenAI, or Gemini (stored encrypted per-user via Windows DPAPI) — or point at a local Ollama / LM Studio server and press Detect models, no key required.
- 5
Press Start
Speech appears as speaker-colored bubbles. Right-click one of your own → This is me to teach it your voice, then let it draft answers, keep notes, or summarize as you go.
Questions or trouble?
Something not working, or a question before you trust it with a meeting? Get in touch — reports from real rooms and real microphones are the most useful kind.