Step-by-step workflow
1
Install faster-whisper
4x faster than original Whisper. Drop-in replacement.
pip install faster-whisper
2
Transcribe with timestamps
Works on any audio file format.
from faster_whisper import WhisperModel
model = WhisperModel("small", device="cpu")
segments, info = model.transcribe("audio.mp3", language=None)
print(f"Detected language: {info.language}")
with open("transcript.txt", "w") as f:
for segment in segments:
line = f"[{segment.start:.1f}s → {segment.end:.1f}s] {segment.text.strip()}"
print(line)
f.write(line + "\n")
3
Generate SRT subtitles
Create subtitle files for any video.
def fmt(s):
h,m,sec,ms = int(s//3600),int((s%3600)//60),int(s%60),int((s%1)*1000)
return f"{h:02d}:{m:02d}:{sec:02d},{ms:03d}"
segments, _ = model.transcribe("video.mp3")
with open("subs.srt","w") as f:
for i,seg in enumerate(segments,1):
f.write(f"{i}\n{fmt(seg.start)} --> {fmt(seg.end)}\n{seg.text.strip()}\n\n")
Pro tips
→
For Hindi: set language="hi", Tamil: "ta", Telugu: "te"
→
large-v3 model = near-perfect accuracy for important meetings
→
Combine with Claude: transcribe → summarise the .txt file
Why this matters for India
// india context
Transcribe legal proceedings, medical consultations, classes — all locally. No data leaves your machine.