Voxta SenseVoice
SenseVoice is a local speech-to-text engine for Voxta that also detects emotion and audio events like laughter, music and applause.
SenseVoice is an open-source neural STT that reports more than words: it detects emotion and audio events — laughter, music, applause. It runs locally.
Setup
Add the service
Manage Modules → Add Modules → SenseVoice → Install. The model downloads on first use.
Pick a model
| Model | Covers |
|---|---|
| SenseVoiceSmall | 5 languages, plus emotions and audio events. |
| Fun-ASR Nano | Chinese, English, Japanese and dialects. |
| Fun-ASR MLT Nano | 31 languages. |
Set the language
Leave it on auto-detect, or pin it if you always speak the same one — pinning is more reliable.
Handles one transcription at a time, and needs some GPU memory.