Function transcribe

funasr/bin/_server_app.py:192–244

(
        file: UploadFile = File(...),
        model: str = Form(default="fun-asr-nano"),
        language: Optional[str] = Form(default=None),
        response_format: Optional[str] = Form(default="json"),
        spk: bool = Form(default=False),
    )

    @app.post("/v1/audio/transcriptions")
    async def transcribe(
        file: UploadFile = File(...),
        model: str = Form(default="fun-asr-nano"),
        language: Optional[str] = Form(default=None),
        response_format: Optional[str] = Form(default="json"),
        spk: bool = Form(default=False),
    ):
        content = await file.read()
        t0 = time.perf_counter()

        if model == "fun-asr-nano":
            _load_vllm_engine()
            if app.state.use_vllm:
                audio_data, sr = sf.read(io.BytesIO(content))
                result = _process_vllm(audio_data, sr, language=language, use_spk=spk)
            else:
                suffix = os.path.splitext(file.filename)[1] if file.filename else ".wav"
                with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
                    tmp.write(content)
                    tmp_path = tmp.name
                try:
                    result = _process_fallback("fun-asr-nano", tmp_path, language=language)
                finally:
                    os.unlink(tmp_path)
        elif model in FALLBACK_CONFIGS:
            suffix = os.path.splitext(file.filename)[1] if file.filename else ".wav"
            with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
                tmp.write(content)
                tmp_path = tmp.name
            try:
                result = _process_fallback(model, tmp_path, language=language)
            finally:
                os.unlink(tmp_path)
        else:
            raise HTTPException(400, f"Unknown model '{model}'. Available: fun-asr-nano, {', '.join(FALLBACK_CONFIGS.keys())}")

        t1 = time.perf_counter()

        if response_format == "verbose_json":
            return JSONResponse({
                "task": "transcribe",
                "language": language or "zh",
                "duration": result.get("duration", 0),
                "text": result["text"],
                "segments": [
                    {"id": i, "start": s["start"], "end": s["end"], "text": s["text"], "words": s.get("words", [])}
                    for i, s in enumerate(result["segments"])
                ],
            })
        elif response_format == "text":
            return JSONResponse(result["text"])
        else:
            return JSONResponse({"text": result["text"]})

    @app.post("/asr")
    async def asr_endpoint(