Voice-in
Transcribe the voice messages you send to Palui with ElevenLabs Scribe, so the agent can answer them like text.
Kind: skill. Voice-in turns the voice messages you send into text, so Palui can answer them like any other message. It uses ElevenLabs Scribe for speech to text.
Voice-in is input only. Palui answers in text. Voice replies (text to speech) are planned, not in v1.
Connect
- Get an API key from ElevenLabs.
- On the Saklar page, add Voice-in and paste the key.
- Press Test.
Settings
| Field | Notes |
|---|---|
| ElevenLabs API key | Secret. Write-only, shown masked. |
Fallback
Gemini audio through CLIProxyAPI is planned as a fallback transcriber. It's conditional: it ships only if testing shows Gemini audio works reliably through CLIProxyAPI. Until then, ElevenLabs Scribe is the only transcriber.
How it behaves
When a voice message arrives on a channel, the runtime sends the audio to the transcriber and puts the transcript into the turn as your message. From there it's a normal turn: same router, same tools, same approvals.
Tools
None. Voice-in works on incoming audio, not as a tool the model calls.
Pause and remove
Pausing applies from the next turn. Removing deletes the stored API key.