@acpfx/tts-kyutai
v0.2.2
Published
Local TTS via Kyutai moshi (on-device, GPU, Python/uv)
Readme
@acpfx/tts-kyutai
Local text-to-speech via Kyutai moshi. Runs on-device with GPU acceleration for high-quality voice synthesis -- no API key needed.
Usage
This package is a pipeline node for @acpfx/cli. See the CLI package for installation and usage.
Uses MLX acceleration on Apple Silicon Macs for fast on-device inference.
Manifest
- Consumes:
agent.delta,agent.complete,agent.tool_start,control.interrupt - Emits:
audio.chunk,lifecycle.ready,lifecycle.done,log
Settings
| Name | Type | Default | Description |
|------|------|---------|-------------|
| model | string | kyutai/tts-1.6b-en_fr | HuggingFace model ID |
| voice | string | expresso/ex03-ex01_happy_001_channel1_334s.wav | Voice name or path to .safetensors file |
| device | string | auto | Compute device: auto, cpu, cuda, or metal |
Pipeline Example
nodes:
tts:
use: "@acpfx/tts-kyutai"
settings: { voice: "expresso/ex03-ex01_happy_001_channel1_334s.wav", device: auto }
outputs: [player]External Links
- Kyutai -- Open-weight AI research lab
- Delayed Streams Modeling -- Kyutai's streaming speech model architecture
License
ISC
