azcomputerguru
a3a47f2d5e
Created Mac M4 batch transcription using mlx-whisper with Apple Silicon GPU acceleration. Transcribed 8 remaining episodes (17,555 total segments). Scripts: - batch_transcribe_mac.py: Full batch processor with mlx-whisper - test_mac_transcribe.py: Quick test script for faster-whisper Transcripts (JSON, SRT, TXT formats): - 2011-06-04-hr1: 1,503 segments - 2011-09-10-hr1: 1,378 segments - 2014-s6e05: 1,340 segments - 2015-s7e30: 1,053 segments - 2016-s8e42: 2,205 segments - 2017-s9e26: 2,366 segments - 2018-s10e17: 4,683 segments - 2018-s10e21: 2,493 segments All 9 episodes now transcribed (8 on Mac + 1 from Linux). Ready for Stages 3-6 on Linux PC. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
76 lines
2.0 KiB
Python
76 lines
2.0 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Quick test script to verify faster-whisper works on Mac M4.
|
|
Transcribes first 60 seconds of an episode.
|
|
"""
|
|
|
|
import time
|
|
from pathlib import Path
|
|
|
|
from faster_whisper import WhisperModel
|
|
from pydub import AudioSegment
|
|
|
|
# Config
|
|
EPISODE = Path("training-data/episodes/2011-06-04-hr1.mp3")
|
|
TEST_DURATION_MS = 60_000 # 60 seconds
|
|
MODEL_SIZE = "base" # Start small for testing, switch to large-v3 for production
|
|
|
|
def main():
|
|
print(f"[INFO] Loading {MODEL_SIZE} model on CPU...")
|
|
start = time.time()
|
|
|
|
# Use CPU - faster-whisper/ctranslate2 doesn't support MPS
|
|
model = WhisperModel(MODEL_SIZE, device="cpu", compute_type="int8")
|
|
print(f"[OK] Model loaded in {time.time() - start:.1f}s")
|
|
|
|
# Extract first 60 seconds
|
|
print(f"[INFO] Extracting first {TEST_DURATION_MS // 1000}s from {EPISODE.name}...")
|
|
audio = AudioSegment.from_mp3(str(EPISODE))
|
|
test_clip = audio[:TEST_DURATION_MS]
|
|
|
|
# Export to temp file
|
|
temp_file = Path("/tmp/test_clip.wav")
|
|
test_clip.export(str(temp_file), format="wav")
|
|
print(f"[OK] Test clip exported ({temp_file.stat().st_size // 1024}KB)")
|
|
|
|
# Transcribe
|
|
print("[INFO] Transcribing...")
|
|
start = time.time()
|
|
|
|
segments, info = model.transcribe(
|
|
str(temp_file),
|
|
language="en",
|
|
beam_size=5,
|
|
vad_filter=True,
|
|
)
|
|
|
|
# Collect segments
|
|
results = []
|
|
for seg in segments:
|
|
results.append({
|
|
"start": seg.start,
|
|
"end": seg.end,
|
|
"text": seg.text.strip()
|
|
})
|
|
|
|
elapsed = time.time() - start
|
|
|
|
print(f"[OK] Transcription complete in {elapsed:.1f}s")
|
|
print(f"[INFO] Speed: {TEST_DURATION_MS / 1000 / elapsed:.2f}x realtime")
|
|
print(f"[INFO] Segments: {len(results)}")
|
|
print()
|
|
print("=" * 60)
|
|
print("TRANSCRIPT:")
|
|
print("=" * 60)
|
|
|
|
for seg in results:
|
|
print(f"[{seg['start']:.1f}s - {seg['end']:.1f}s] {seg['text']}")
|
|
|
|
# Cleanup
|
|
temp_file.unlink()
|
|
print()
|
|
print("[SUCCESS] Test complete!")
|
|
|
|
if __name__ == "__main__":
|
|
main()
|