#!/usr/bin/env python3 """Generate C1 (concat) and C2 (crossfade) MixSet from ACE-Step + MTG.""" import json, os, random, subprocess, time from pathlib import Path ACESTEP_DIR = Path("/ssd_data/dataset/haim_dataset/fake/acestep/samples") MTG_DIR = Path("/ssd_data/dataset/haim_dataset/real/MTG") C1_DIR = Path("/ssd_data/dataset/haim_dataset/C_mixing/C1_mixset_concat") C2_DIR = Path("/ssd_data/dataset/haim_dataset/C_mixing/C2_mixset_crossfade") TARGET = 6000 def get_duration(path): r = subprocess.run(['ffprobe','-v','quiet','-show_entries','format=duration', '-of','csv=p=0', str(path)], capture_output=True, text=True, timeout=5) return float(r.stdout.strip()) if r.stdout.strip() else 0 def get_segment(path, start, duration, out_path): """Extract segment from audio file.""" subprocess.run([ 'ffmpeg','-y','-i',str(path),'-ss',str(start),'-t',str(duration), '-ar','44100','-ac','2','-ab','192k', str(out_path) ], capture_output=True, timeout=60) def concat_with_gap(seg1, seg2, gap, out_path): """Concatenate two segments with silent gap.""" subprocess.run([ 'ffmpeg','-y','-i',str(seg1),'-i',str(seg2), '-filter_complex', f'[0:a]apad=pad_dur={gap}[a0];[a0][1:a]concat=n=2:v=0:a=1[out]', '-map','[out]','-ar','44100','-ac','2','-ab','192k', str(out_path) ], capture_output=True, timeout=60) def crossfade_mix(seg1, seg2, xfade_dur, out_path): """Crossfade two segments.""" subprocess.run([ 'ffmpeg','-y','-i',str(seg1),'-i',str(seg2), '-filter_complex', f'[0:a][1:a]acrossfade=d={xfade_dur}:c1=tri:c2=tri[out]', '-map','[out]','-ar','44100','-ac','2','-ab','192k', str(out_path) ], capture_output=True, timeout=60) def main(): random.seed(42) ai_files = sorted(ACESTEP_DIR.glob("*.mp3")) human_files = sorted(MTG_DIR.glob("*.mp3")) print(f"AI: {len(ai_files)}, Human: {len(human_files)}") C1_DIR.mkdir(parents=True, exist_ok=True) C2_DIR.mkdir(parents=True, exist_ok=True) # Shuffle and pair ai_pool = list(ai_files) human_pool = list(human_files) random.shuffle(ai_pool) random.shuffle(human_pool) tmp_dir = Path("/tmp/mixset_tmp") tmp_dir.mkdir(exist_ok=True) for i in range(TARGET): c1_mp3 = C1_DIR / f"C1_{i:05d}.mp3" c2_mp3 = C2_DIR / f"C2_{i:05d}.mp3" if c1_mp3.exists() and c2_mp3.exists(): continue ai_src = ai_pool[i % len(ai_pool)] human_src = human_pool[i % len(human_pool)] try: ai_dur = get_duration(ai_src) human_dur = get_duration(human_src) if ai_dur < 20 or human_dur < 20: continue # Random segment lengths (15-60s) ai_seg_len = min(random.uniform(15, 60), ai_dur - 1) human_seg_len = min(random.uniform(15, 60), human_dur - 1) ai_start = random.uniform(0, max(0, ai_dur - ai_seg_len)) human_start = random.uniform(0, max(0, human_dur - human_seg_len)) # Extract segments seg_ai = tmp_dir / f"ai_{i}.mp3" seg_human = tmp_dir / f"human_{i}.mp3" get_segment(ai_src, ai_start, ai_seg_len, seg_ai) get_segment(human_src, human_start, human_seg_len, seg_human) if not seg_ai.exists() or not seg_human.exists(): continue # Randomize order (AI first or Human first) ai_first = random.choice([True, False]) seg1 = seg_ai if ai_first else seg_human seg2 = seg_human if ai_first else seg_ai order = "ai_first" if ai_first else "human_first" # C1: Concat with gap gap = round(random.uniform(0.1, 0.5), 2) if not c1_mp3.exists(): concat_with_gap(seg1, seg2, gap, c1_mp3) seg1_dur = get_duration(seg1) boundary = round(seg1_dur + gap, 3) meta_c1 = { "track_id": f"C1_{i:05d}", "filename": c1_mp3.name, "method": "concat", "order": order, "gap_sec": gap, "segment_1": { "source": ai_src.name if ai_first else human_src.name, "type": "ai" if ai_first else "human", "start": round(ai_start if ai_first else human_start, 2), "duration": round(ai_seg_len if ai_first else human_seg_len, 2), }, "segment_2": { "source": human_src.name if ai_first else ai_src.name, "type": "human" if ai_first else "ai", "start": round(human_start if ai_first else ai_start, 2), "duration": round(human_seg_len if ai_first else ai_seg_len, 2), }, "boundary_sec": boundary, } with open(C1_DIR / f"C1_{i:05d}.json", "w", encoding="utf-8") as f: json.dump(meta_c1, f, ensure_ascii=False, indent=2) # C2: Crossfade xfade = round(random.uniform(1, 5), 2) if not c2_mp3.exists(): crossfade_mix(seg1, seg2, xfade, c2_mp3) seg1_dur = get_duration(seg1) boundary_start = round(seg1_dur - xfade, 3) boundary_end = round(seg1_dur, 3) meta_c2 = { "track_id": f"C2_{i:05d}", "filename": c2_mp3.name, "method": "crossfade", "order": order, "crossfade_sec": xfade, "segment_1": { "source": ai_src.name if ai_first else human_src.name, "type": "ai" if ai_first else "human", "start": round(ai_start if ai_first else human_start, 2), "duration": round(ai_seg_len if ai_first else human_seg_len, 2), }, "segment_2": { "source": human_src.name if ai_first else ai_src.name, "type": "human" if ai_first else "ai", "start": round(human_start if ai_first else ai_start, 2), "duration": round(human_seg_len if ai_first else ai_seg_len, 2), }, "boundary_start_sec": boundary_start, "boundary_end_sec": boundary_end, } with open(C2_DIR / f"C2_{i:05d}.json", "w", encoding="utf-8") as f: json.dump(meta_c2, f, ensure_ascii=False, indent=2) # Cleanup tmp seg_ai.unlink(missing_ok=True) seg_human.unlink(missing_ok=True) if (i + 1) % 100 == 0: c1_count = len(list(C1_DIR.glob("*.mp3"))) c2_count = len(list(C2_DIR.glob("*.mp3"))) print(f"[{i+1}/{TARGET}] C1={c1_count} C2={c2_count}") except Exception as e: print(f"[{i}] Error: {e}") continue c1_final = len(list(C1_DIR.glob("*.mp3"))) c2_final = len(list(C2_DIR.glob("*.mp3"))) print(f"Done: C1={c1_final}, C2={c2_final}") if __name__ == "__main__": main()