Voice & Speech
Anyy separates incoming audio transcription from spoken replies:
- Speech-to-Text (STT) turns an audio attachment into text for the conversation.
- Text-to-Speech (TTS) creates an audio reply from Anyy's response.
You can configure either capability without enabling the other.
Configure in Dashboard
Open Dashboard → Tools → Voice.
For TTS, choose the provider/protocol, model, voice, endpoint, response format, timeout, and maximum input length. Use the preview action before enabling automatic voice replies.
For STT, choose the provider, model, optional language, endpoint, and timeout, then run verification with a supported audio sample or provider probe.
CLI wizards are also available:
anyy tools voice tts
anyy tools voice stt
Supported Route Types
Speech routes can use supported OpenAI or OpenAI-compatible endpoints. TTS also supports the configured chat-audio or local command route; STT can use a configured command route. A command route executes the exact trusted command configured for the Profile—do not point it at unreviewed programs.
Transcribe Audio
Attach a supported audio or voice file and ask Anyy to transcribe or answer it. The file is staged within the session's attachment scope, then sent to the configured STT route when transcription is required.
Adding a language hint can improve recognition when the provider supports it. Do not set the wrong language for multilingual audio.
Generate Spoken Replies
Ask explicitly:
Reply with a short voice message in a calm tone.
When TTS succeeds, Anyy returns an audio attachment. Delivery depends on the current surface and channel's supported media kinds.
Automatic Voice Replies
The Profile can enable automatic TTS. In supported channel conversations, /voice manages voice replies for that conversation. Weixin exposes /voice together with its other channel commands.
Audio, transcript text, or reply text may leave the device. Review provider privacy, retention, regional, and billing terms before using sensitive recordings.
Troubleshooting
| Symptom | Check |
|---|---|
| Audio is attached but not transcribed | Verify STT is enabled, the format is supported, and the route passes its test |
| TTS tool is unavailable | Enable TTS and confirm the tts toolset is available |
| Voice preview fails | Check endpoint, credential, model, voice name, and response format |
| Channel receives text only | Confirm auto TTS/conversation voice mode and the channel's audio delivery support |
| Speech request is cut off | Review maximum characters, attachment size, and timeout |