Apppublic

SalehAhmed10/mbts-tts

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

0likes

App README

MBTS TTS Service

HeadTTS Node server running Kokoro-82M (q8 quantized) ONNX inference. Returns audio + Oculus viseme timestamps for lip-sync animation.

API

`GET /v1/hello`

Health check. Returns "HeadTTS-Node v1.3.0"

`POST /v1/synthesize`

Generate speech with viseme data.

Request:

json

{
  "input": "Hello, welcome to MBTS.",
  "voice": "af_bella",
  "language": "en-us",
  "speed": 1,
  "audioEncoding": "wav"
}

Response:

json

{
  "audio": "<base64-encoded-wav-data>",
  "words": ["Hello, ", "welcome ", "to ", "MBTS."],
  "wtimes": [100, 450, 800, 1000],
  "wdurations": [350, 350, 200, 600],
  "visemes": ["aa", "E", "SS", "I", "SS", "TH", "nn", "DD"],
  "vtimes": [100, 200, 300, 450, 550, 650, 750, 850],
  "vdurations": [100, 100, 150, 100, 100, 100, 100, 150],
  "phonemes": ["ð", "ɪ", "s", "ɪ", "z"],
  "audioEncoding": "wav"
}

Voices

—af_bella — American female (default)
—am_fenrir — American male
—ef_dora — British female

Max input length

500 characters per request.

Built With

—HeadTTS — Kokoro TTS with viseme timestamps
—Kokoro-82M — ONNX timestamped model
—Hugging Face Spaces (CPU Basic, free tier)