[15:53:26] === Mac voice-extract: idol=winter, urls=3 === [15:53:26] root: /Users/inhohur/ai/experiments/voice-extract-mac/winter/data [15:53:26] device: mps [15:53:26] [1/7] fetch [15:53:26] fetch [skip] de6FhGIFoxA [15:53:26] fetch [skip] OrFTzGP57MQ [15:53:26] fetch [skip] yOIZ53fviVk [15:53:26] [2/7] vocals (demucs-mlx) [15:53:26] vocals [skip] de6FhGIFoxA [15:53:26] vocals [skip] OrFTzGP57MQ [15:53:26] vocals [skip] yOIZ53fviVk [15:53:26] [3/7] diarize (pyannote 3.1 / MPS) [15:53:30] diarize [skip] de6FhGIFoxA [15:53:30] diarize [skip] OrFTzGP57MQ [15:53:30] diarize [skip] yOIZ53fviVk [15:53:30] [4/7] select [15:53:30] select auto_dominant: SPEAKER_01 (total 1627.0s) [15:53:30] SPEAKER_01: 1627.0s [15:53:30] SPEAKER_00: 122.9s [15:53:30] SPEAKER_02: 45.8s [15:53:30] SPEAKER_03: 25.3s [15:53:30] [5/8] match (WeSpeaker ResNet34 / MPS) [15:53:30] match [load WeSpeaker] on mps ... [15:53:31] loaded (1.0s) [15:53:31] match [centroid] from de6FhGIFoxA SPEAKER_01 ... [15:53:54] centroid: 648 segments [15:54:01] match [de6FhGIFoxA] 497/1430 segments ≥ 0.55 (6.3s) [15:54:08] match [OrFTzGP57MQ] 343/1100 segments ≥ 0.55 (7.3s) [15:54:11] match [yOIZ53fviVk] 194/548 segments ≥ 0.55 (3.0s) [15:54:11] matched total: 1034 [15:54:11] [6/8] stitch (인접 matched → 10-15s clips) [15:54:11] stitch [de6FhGIFoxA]: 2 clips (10.0-15.0s) [15:54:11] stitch [OrFTzGP57MQ]: 1 clips (10.0-15.0s) [15:54:12] stitch [yOIZ53fviVk]: 1 clips (10.0-15.0s) [15:54:12] stitched total: 4 [15:54:12] [7/8] filter (VAD + SNR) [15:54:12] filter [silero-vad] loading ... [15:54:12] filter kept: 2 / 4 stitched clips [15:54:12] rejected: {'no_voice': 0, 'voice_dur': 0, 'voice_ratio': 1, 'snr': 1} [15:54:12] [8/8] concat + transcribe [15:54:12] prompt prompt.wav: 13.42s (de6FhGIFoxA_04609420_13415_0.794.wav, cos=0.794) [15:54:12] transcribe [docksvc whisper] prompt.wav ... [15:54:13] done (1.7s) — 87 chars [15:54:13] "학교 다닐 때 봉사 시간을 채워야 했단 말이에요 그 무슨 무조건 채워야 되는 모두가 이제 모두가 봉사 시간 10시간인가 무조건 그게 수행평가에요 수행평가였어" [15:54:13] === done === [15:54:13] dataset_dir: /Users/inhohur/ai/experiments/voice-extract-mac/winter/data/filtered [15:54:13] prompt.wav: /Users/inhohur/ai/experiments/voice-extract-mac/winter/data/prompt.wav [15:54:13] prompt.txt: /Users/inhohur/ai/experiments/voice-extract-mac/winter/data/prompt.txt [15:54:13] manifest: /Users/inhohur/ai/experiments/voice-extract-mac/winter/data/manifest.json