62 lines
1.9 KiB
Python
62 lines
1.9 KiB
Python
import os
|
||
import tempfile
|
||
from typing import Optional
|
||
|
||
from fastapi import FastAPI, UploadFile, File, Form
|
||
from fastapi.responses import JSONResponse
|
||
from faster_whisper import WhisperModel
|
||
|
||
MODEL_NAME = os.getenv("FW_MODEL", "large-v3")
|
||
DEVICE = os.getenv("FW_DEVICE", "cuda")
|
||
COMPUTE_TYPE = os.getenv("FW_COMPUTE_TYPE", "float16")
|
||
|
||
app = FastAPI()
|
||
|
||
# 起動時にモデルをロードして常駐（ここが重要：毎回ロードしない）
|
||
model = WhisperModel(MODEL_NAME, device=DEVICE, compute_type=COMPUTE_TYPE)
|
||
|
||
@app.get("/health")
|
||
def health():
|
||
return {"status": "ok", "model": MODEL_NAME, "device": DEVICE, "compute_type": COMPUTE_TYPE}
|
||
|
||
@app.post("/transcribe")
|
||
async def transcribe(
|
||
file: UploadFile = File(...),
|
||
language: Optional[str] = Form(default=None), # "ja"など
|
||
beam_size: int = Form(default=5),
|
||
vad_filter: bool = Form(default=True),
|
||
):
|
||
# アップロードを一時ファイルに保存（PyAVがファイルパスを扱いやすい）
|
||
suffix = os.path.splitext(file.filename or "audio.wav")[1] or ".wav"
|
||
with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
|
||
tmp_path = tmp.name
|
||
tmp.write(await file.read())
|
||
|
||
try:
|
||
segments, info = model.transcribe(
|
||
tmp_path,
|
||
language=language,
|
||
beam_size=beam_size,
|
||
vad_filter=vad_filter,
|
||
)
|
||
|
||
out_segments = []
|
||
text_all = []
|
||
for s in segments:
|
||
out_segments.append({"start": s.start, "end": s.end, "text": s.text})
|
||
text_all.append(s.text)
|
||
|
||
return JSONResponse(
|
||
{
|
||
"language": info.language,
|
||
"language_probability": info.language_probability,
|
||
"text": "".join(text_all).strip(),
|
||
"segments": out_segments,
|
||
}
|
||
)
|
||
finally:
|
||
try:
|
||
os.remove(tmp_path)
|
||
except OSError:
|
||
pass
|