[14:17:14] === Mac voice-extract: idol=taeyeon, urls=3 === [14:17:14] root: /Users/inhohur/ai/experiments/voice-extract-mac/taeyeon/data [14:17:14] device: mps [14:17:14] [1/7] fetch [14:17:14] fetch [yt-dlp] 8k_sSfbX7rY ... [14:17:19] done (719s, 4.8s elapsed) [14:17:19] fetch [yt-dlp] iBX5T-q9bcI ... [14:17:22] done (209s, 2.9s elapsed) [14:17:22] fetch [yt-dlp] 6dpLwoCW4Lw ... [14:17:27] done (401s, 4.7s elapsed) [14:17:27] [2/7] vocals (demucs-mlx) [14:17:27] vocals [demucs-mlx] batch separating 3 track(s) ... [14:17:48] batch done (21.1s elapsed) [14:17:48] [8k_sSfbX7rY] -> 8k_sSfbX7rY.wav [14:17:48] [iBX5T-q9bcI] -> iBX5T-q9bcI.wav [14:17:49] [6dpLwoCW4Lw] -> 6dpLwoCW4Lw.wav [14:17:49] [3/7] diarize (pyannote 3.1 / MPS) [14:17:53] diarize [pyannote] loading speaker-diarization-3.1 on mps ... [14:17:55] loaded (1.7s) [14:17:55] diarize [run] 8k_sSfbX7rY ... [14:18:13] done (18.3s) — 264 segments, 2 speakers: ['SPEAKER_00', 'SPEAKER_01'] [14:18:13] diarize [run] iBX5T-q9bcI ... [14:18:18] done (4.9s) — 58 segments, 1 speakers: ['SPEAKER_00'] [14:18:18] diarize [run] 6dpLwoCW4Lw ... [14:18:27] done (9.3s) — 85 segments, 2 speakers: ['SPEAKER_00', 'SPEAKER_01'] [14:18:27] [4/7] select [14:18:27] select auto_dominant: SPEAKER_01 (total 448.8s) [14:18:27] SPEAKER_01: 448.8s [14:18:27] SPEAKER_00: 8.4s [14:18:27] [5/7] match (WeSpeaker ResNet34 / MPS) [14:18:27] match [load WeSpeaker] on mps ... [14:18:27] loaded (0.4s) [14:18:27] match [centroid] from 8k_sSfbX7rY SPEAKER_01 ... [14:18:42] centroid: 152 segments [14:18:43] match [8k_sSfbX7rY] 82/264 segments ≥ 0.55 (1.4s) [14:18:48] match [iBX5T-q9bcI] 3/58 segments ≥ 0.55 (4.3s) [14:19:00] match [6dpLwoCW4Lw] 23/85 segments ≥ 0.55 (12.2s) [14:19:00] matched total: 108 [14:19:00] [6/7] filter [14:19:00] filter [silero-vad] loading ... [14:19:01] filter kept: 63 segments [14:19:01] [7/7] concat + transcribe [14:19:01] concat prompt.wav: 30.00s (5 chunks) [14:19:01] transcribe [docksvc whisper] prompt.wav ... [14:19:05] done (3.7s) — 249 chars [14:19:05] "내가 예전에 이거 뭐라고 썼는지 기억이 나는데 뭐 약간 뭐라.. 뭐라.. 되게 애매하게 얘기했는데 느끼하지 않은 정도? 뭐 이렇게 썼나? 뭐 이랬을 거야 밑에 춤 솜씨도 기억나 어색하지 않은 정도 뭐 이렇게 썼을 거야 그 멤버가 남긴 멘트 아 뭐라고 했냐면요 98번 질문이야 벌써 기억에 남는 주접 댓글 지구가 반으로 접혀서 나 어디야? 막 이런 거 있잖아요 말도 안 되는 진짜 그런 거 있었네 생각을 갖고 있고 어떤 정보가 업데이트 됐는지 한번" [14:19:05] === done === [14:19:05] dataset_dir: /Users/inhohur/ai/experiments/voice-extract-mac/taeyeon/data/filtered [14:19:05] prompt.wav: /Users/inhohur/ai/experiments/voice-extract-mac/taeyeon/data/prompt.wav [14:19:05] prompt.txt: /Users/inhohur/ai/experiments/voice-extract-mac/taeyeon/data/prompt.txt [14:19:05] manifest: /Users/inhohur/ai/experiments/voice-extract-mac/taeyeon/data/manifest.json