FUNASR DOCUMENTATION

From first transcript to production.

Model capabilities, runnable examples, and deployment boundaries in one place.

01

Get started

Transformers quickstart Installation & environment checks Your first transcription Python SDK: AutoModel Choose your first model Command-line transcription Migrate from Whisper & cloud APIs

02

Models & capabilities

MOSS transcription & diarization Model Zoo Streaming voice activity detection Keyword spotting Speakers and emotion tags

03

Train & extend

Fine-tuning & evaluation Custom model registration

04

Deploy & integrate

Containers & image selection Serving entry points Python HTTP service Kubernetes deployment Runtime & deployment matrix vLLM split-engine inference Official native vLLM functional validation (2026-09-07) Historical community native vLLM record (2026-08-13)

05

Evaluate & operate

Troubleshooting Service security boundaries Reproducible performance tests Historical ASR benchmark Realtime WebSocket benchmarking

06

Applications & ecosystem

Agent integration Use cases & examples Community integrations Repositories & contribution paths JavaScript clients Gradio browser client Workflow integrations

07

Protocols & reference

HTTP API contract C++ WebSocket protocol ONNX binary output contract