# OmniVoice Gateway

> High-performance voice processing gateway for text-to-speech (TTS) voice cloning, automatic speech recognition (ASR), and video dubbing.

## Documentation and Specifications

- [API Documentation](/docs.md): Comprehensive Markdown reference of all API endpoints, models, parameters, and implementation code.
- [OpenAPI Specification](/v1/openapi.json): Raw OpenAPI JSON spec file served directly from the backend.

## Core Features

- **Text-to-Speech (TTS)**: Multi-mode generation (`clone_voice`, `auto_voice`, `voice_design`) with neural inference control (`speed`, `num_step`, `guidance_scale`).
- **Word-Level Alignment**: High-precision timestamps (start, end) for every generated word, ideal for karaoke-style subtitles.
- **Fine-Grained Controls**: Expressions (`[sigh]`, `[laughter]`) and phonetic/Pinyin overrides for custom pronunciation correction.
- **Voice Library & Management**: Publish, tag, search, and update voice samples.
- **ElevenLabs Compatibility**: Native compat routes mapping ElevenLabs client libraries to the OmniVoice engine.
- **Speech-to-Text (ASR)**: Job-based audio transcription pipeline.
- **Video Dubbing**: Orchestrated video dubbing pipeline (YouTube/MP4 download -> vocal separation -> transcription -> translation -> TTS dubbing -> BGM remix).