livekit-streaming-tts · control panel

🩺 Health & Capabilities GET /health

Live status of the running engine. Auto-refreshes every 5s.

status

—

capabilities

—

auto

🎙️ Voices GET /v1/voices

Registered + built-in voices. Click use to load into the TTS panels.

—

🧬 Clone a Voice POST /v1/voices/upload

Upload a reference clip to register a cloned voice. Requires a cloning-capable engine.

Voice ID Reference audio (wav / mp3 / flac)

Reference transcript (optional, some engines need it)

🔊 Test TTS (one-shot) POST /v1/tts

Synthesize the whole utterance in one HTTP request and play it back.

Text

Voice

Language

Format

Speed

Seed

Sample rate

normalize text (numbers, money, times, diacritics)

Off — your text reaches the model exactly as typed, so your own tashkeel survives. Digits, times and currency are not expanded: write them as words.

Advanced — engine-specific extras

extra (JSON) e.g. {"temperature":0.7, "steps":8}

🌊 Streaming TTS (WebSocket) WS /tts/ws

Full duplex protocol: config → text → flush. Supports barge-in (cancel) and ping. Audio plays live as each chunk arrives; the full utterance is also collected below for replay.

Voice

Language

Speed

Seed

Format

Sample rate

binary PCM frames (format must be pcm)

normalize text (numbers, money, times, diacritics)

Off — text passes through untouched, keeping your tashkeel. Digits are not expanded. Takes effect immediately on a live session.

play live (speak each chunk on arrival)

Connection

Text to stream

Protocol log

Live playback

idle

Collected audio (replay / download)

No chunks yet.

Served from the FastAPI app · same-origin so WebSocket & uploads work without CORS config.