Spaces:
Sleeping
Sleeping
File size: 5,085 Bytes
5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d a8277c5 740950d 5aea4af d7ee472 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d 5aea4af 740950d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 | import os
import shutil
import tempfile
from typing import Any, Dict, List, Optional
from fastapi import ( # type: ignore[import]
FastAPI,
File,
UploadFile,
HTTPException,
)
from fastapi.responses import JSONResponse # type: ignore[import]
try:
# Import lazily; model will be instantiated on first use
from faster_whisper import WhisperModel # type: ignore
except Exception as import_error: # pragma: no cover
WhisperModel = None # type: ignore
_import_error: Optional[Exception] = import_error
else:
_import_error = None
app = FastAPI(title="faster-whisper REST API")
def _get_bool_env(name: str, default: bool = False) -> bool:
val = os.getenv(name)
if val is None:
return default
return val.strip().lower() in {"1", "true", "yes", "y", "on"}
class _ModelHolder:
_model: Optional[WhisperModel] = None # type: ignore
@classmethod
def get_model(cls) -> "WhisperModel": # type: ignore
if _import_error is not None:
raise RuntimeError(
f"Failed to import faster-whisper: {_import_error}"
)
if cls._model is None:
model_size = os.getenv("WHISPER_MODEL", "small")
device = os.getenv("WHISPER_DEVICE", "cpu")
compute_type = os.getenv("WHISPER_COMPUTE_TYPE", "int8")
# Construct model once and reuse across requests
cls._model = WhisperModel(
model_size,
device=device,
compute_type=compute_type,
) # type: ignore
return cls._model
@app.get("/")
def root() -> Dict[str, Any]:
return {"status": "ok"}
@app.get("/health")
def health() -> Dict[str, Any]:
return {"status": "ok"}
@app.post("/inference")
def transcribe(
file: UploadFile = File(...),
language: Optional[str] = None,
beam_size: Optional[int] = None,
vad: Optional[bool] = None,
word_timestamps: Optional[bool] = None,
) -> JSONResponse:
try:
model = _ModelHolder.get_model()
except Exception as e: # pragma: no cover
raise HTTPException(status_code=500, detail=str(e))
# Fallback to env-configurable defaults
effective_beam_size = (
int(os.getenv("WHISPER_BEAM_SIZE", "5"))
if beam_size is None
else beam_size
)
effective_vad = (
_get_bool_env("WHISPER_VAD", False)
if vad is None
else vad
)
effective_word_ts = (
_get_bool_env("WHISPER_WORD_TIMESTAMPS", False)
if word_timestamps is None
else word_timestamps
)
# Persist upload to a temporary file so ffmpeg/backends can access it
try:
with tempfile.NamedTemporaryFile(
delete=False,
suffix=os.path.splitext(file.filename or "audio")[1],
) as tmp:
shutil.copyfileobj(file.file, tmp)
tmp_path = tmp.name
except Exception as e:
raise HTTPException(
status_code=400,
detail=f"Failed to read uploaded file: {e}",
)
finally:
try:
file.file.close()
except Exception:
pass
try:
segments_iter, info = model.transcribe(
tmp_path,
beam_size=effective_beam_size,
language=language,
vad_filter=effective_vad,
word_timestamps=effective_word_ts,
)
segments = list(segments_iter)
# Build response
all_text = "".join(seg.text for seg in segments)
segments_json: List[Dict[str, Any]] = []
for seg in segments:
seg_obj: Dict[str, Any] = {
"id": seg.id,
"start": seg.start,
"end": seg.end,
"text": seg.text,
"avg_log_prob": seg.avg_logprob,
"compression_ratio": seg.compression_ratio,
"no_speech_prob": seg.no_speech_prob,
}
if effective_word_ts and getattr(seg, "words", None):
seg_obj["words"] = [
{
"start": w.start,
"end": w.end,
"word": w.word,
"probability": getattr(w, "probability", None),
}
for w in seg.words
]
segments_json.append(seg_obj)
result: Dict[str, Any] = {
"text": all_text.strip(),
"segments": segments_json,
"info": {
"language": getattr(info, "language", None),
"language_probability": getattr(
info, "language_probability", None
),
"duration": getattr(info, "duration", None),
},
}
return JSONResponse(result)
except Exception as e:
raise HTTPException(
status_code=500,
detail=f"Transcription error: {e}",
)
finally:
try:
os.unlink(tmp_path)
except Exception:
pass
|