Spaces:
Sleeping
Sleeping
File size: 1,951 Bytes
d309047 | 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 | """
Speech-to-text using local faster-whisper.
Accepts raw audio bytes (webm/wav) and returns transcript with timestamps.
"""
from faster_whisper import WhisperModel
import os
import tempfile
import logging
from typing import Optional
from faster_whisper import WhisperModel
try:
import imageio_ffmpeg
ffmpeg_dir = os.path.dirname(imageio_ffmpeg.get_ffmpeg_exe())
os.environ["PATH"] += os.pathsep + ffmpeg_dir
except Exception as e:
logging.warning(f"Could not load imageio_ffmpeg: {e}")
_model: Optional[WhisperModel] = None
def load_stt_model() -> WhisperModel:
global _model
if _model is None:
print("Loading WhisperModel (base.en)...")
# Run on CPU for broad hackathon compatibility, int8 precision
_model = WhisperModel("base.en", device="cpu", compute_type="int8")
return _model
async def transcribe_audio(audio_bytes: bytes, audio_format: str = "webm") -> dict:
"""
Transcribe audio bytes using local faster-whisper.
Returns { text, segments: [{ text, start, end, speaker }] }
"""
model = load_stt_model()
with tempfile.NamedTemporaryFile(suffix=f".{audio_format}", delete=False) as tmp:
tmp.write(audio_bytes)
tmp_path = tmp.name
try:
segments_gen, info = model.transcribe(tmp_path, beam_size=5, vad_filter=True)
segments = []
full_text = ""
for seg in segments_gen:
segments.append({
"text": seg.text.strip(),
"start": seg.start,
"end": seg.end,
"speaker": "Speaker",
})
full_text += seg.text + " "
if not segments:
segments = [{
"text": "",
"start": 0.0,
"end": 0.0,
"speaker": "Speaker",
}]
return {"text": full_text.strip(), "segments": segments}
finally:
os.unlink(tmp_path)
|