Voice
Dictate a message and have replies read aloud, against any OpenAI-compatible speech servers you run.
Set it up
Admin → Audio takes two endpoints, because they are usually two servers:
| Speaks | For example | |
|---|---|---|
| Dictation | POST /v1/audio/transcriptions | whisper.cpp’s whisper-server, Speaches, faster-whisper-server |
| Read aloud | POST /v1/audio/speech | Kokoro-FastAPI, OpenAI |
If the speech server also answers GET /v1/audio/voices, the voice list is read from it, and each person picks their own voice and speed under Settings → Audio.
HTTPS required. Browsers only allow the microphone over HTTPS or on localhost. An install on a LAN address over plain http will not offer dictation.
Privacy
Recorded audio is passed straight through to your transcription server and never written to disk. The speech servers’ keys never leave the server.
Permissions
Dictation and read-aloud are separate permissions, so a group can have one without the other.
For devices and programs
The same two endpoints are on the API, so a LLeMbas CLI logged in to the server needs no speech server of its own: /v1/audio/transcriptions uses your language, /v1/audio/speech your voice and speed, unless the request says otherwise. See API.
In the CLI
Press Ctrl+T to record and again (or stop talking) to transcribe. /speak reads replies aloud. The voice: block configures it — see CLI configuration.