24 lines
790 B
Markdown
24 lines
790 B
Markdown
# Hermes Qwen3-TTS PCM streaming adapter
|
|
|
|
This optional Hermes backend plugin uses Athena's native
|
|
`/v1/audio/speech/pcm-stream` route. It starts playback while Qwen3-TTS is
|
|
still synthesizing the current sentence instead of waiting for a complete
|
|
audio file.
|
|
|
|
Install this directory as `${HERMES_HOME}/plugins/qwen3-stream`, enable the
|
|
plugin and set:
|
|
|
|
```yaml
|
|
tts:
|
|
provider: qwen3-stream
|
|
streaming:
|
|
provider: qwen3-stream
|
|
```
|
|
|
|
The adapter reuses `tts.openai.base_url`, `tts.openai.api_key`, model, voice
|
|
and language unless an explicit `tts.qwen3-stream` section overrides them.
|
|
This avoids copying the Athena credential into another file.
|
|
|
|
Rollback is immediate: restore `tts.provider` and `tts.streaming.provider` to
|
|
`openai`, disable the plugin and restart the Hermes gateway.
|