# -*- coding: utf-8 -*- """Generate V2 subtitle files with correct timing matching voiceover segments.""" import json import os # V2 segment boundaries from actual voiceover durations seg_boundaries = [ (0.0, 8.93), # seg_01_title (8.93, 18.17), # seg_02_recording_fail (18.17, 26.43), # seg_03_ai_drift (26.43, 36.61), # seg_04_ui_render (36.61, 49.09), # seg_05_kimi_score (49.09, 58.81), # seg_06_v10_crash (58.81, 69.03), # seg_07_rollback (69.03, 80.5), # seg_08_deploy (80.5, 91.37), # seg_09_ending ] # Each segment has list of chunks, each chunk is a list of display lines (max 2, max 13 chars each) seg_chunks_nested = [ # Seg 1 (0-8.93s) [ ["我让 Claude Code", "自己做视频"], ["录屏失败", "AI 跑偏"], ["从 69 爬到 92", "这是真实工作流"], ], # Seg 2 (8.93-18.17s) [ ["想录下执行过程", "但 Windows 抓不到后台桌面"], ["五种方案全是空文件", "录屏彻底失败"], ], # Seg 3 (18.17-26.43s) [ ["更糟的是", "Claude Code 跑偏了"], ["跑去研究亚马逊", "ChatGPT 说：别录了"], ["让它自己画界面"], ], # Seg 4 (26.43-36.61s) [ ["于是换条路", "Python 画终端和工具卡片"], ["四个 AI 并行跑", "进度条在涨"], ["每条命令来自真实日志"], ], # Seg 5 (36.61-49.09s) [ ["每轮渲染完", "Kimi 审核打分"], ["v1 只有 69", "改字体到 72"], ["加动画跳到 81", "v4 85，v6 89"], ["v9 冲到 92"], ], # Seg 6 (49.09-58.81s) [ ["V9 冲到 92", "马上要发布"], ["但下轮硬编码没改", "画面四个代理底部显示三个"], ["92 砸到 76"], ], # Seg 7 (58.81-69.03s) [ ["下轮修了六个问题", "回到 88"], ["但回不到 92 了", "最终回滚 V9 冻结"], ["一个硬编码 bug", "一夜回到解放前"], ], # Seg 8 (69.03-80.5s) [ ["V9 上线审核门户", "HTTP 200 通过"], ["11 轮迭代 1 次崩溃", "69 到 92"], ["AI 自己审核", "修复、部署"], ], # Seg 9 (80.5-91.37s) [ ["录屏失败", "AI 跑偏"], ["分数崩过", "又爬回来"], ["这才是真正的 AI 工作流"], ["执行、审核、修复", "回归、交付"], ], ] # Keywords to highlight in yellow HIGHLIGHT_KEYWORDS = [ "录屏失败", "AI 跑偏", "真实日志", "Kimi 审核", "69", "92", "76", "88", "81", "85", "89", "72", "回滚", "冻结", "HTTP 200", "审核门户", "Claude Code", "ChatGPT", "bug", ] output_dir = "E:/集群文件夹/factory_os/short_video_real_data_pipeline/phase4b_90s_formal_sample/v22_kimi_claude_loop/full_story_douyin_video/04_subtitles/v2" # Build flat list of chunks with times chunks_with_time = [] for seg_idx, (seg_start, seg_end) in enumerate(seg_boundaries): chunks_list = seg_chunks_nested[seg_idx] seg_dur = seg_end - seg_start num_chunks = len(chunks_list) chunk_dur = seg_dur / num_chunks for ci, chunk_lines in enumerate(chunks_list): chunk_start = seg_start + ci * chunk_dur chunk_end = chunk_start + chunk_dur chunks_with_time.append((chunk_start, chunk_end, chunk_lines)) print(f"Total subtitle chunks: {len(chunks_with_time)}") for i, (s, e, lines) in enumerate(chunks_with_time): display = " | ".join(lines) print(f" {i+1}. {s:.1f}s-{e:.1f}s: {display}") # --- Helper functions --- def to_srt_time(seconds): h = int(seconds // 3600) m = int((seconds % 3600) // 60) s = int(seconds % 60) ms = int((seconds - int(seconds)) * 1000) return f"{h:02d}:{m:02d}:{s:02d},{ms:03d}" def to_ass_time(seconds): h = int(seconds // 3600) m = int((seconds % 3600) // 60) s = seconds % 60 return f"{h}:{m:02d}:{s:05.2f}" # --- Generate SRT --- srt_lines = [] for i, (start, end, lines) in enumerate(chunks_with_time, 1): srt_lines.append(str(i)) srt_lines.append(f"{to_srt_time(start)} --> {to_srt_time(end)}") srt_lines.append("\n".join(lines)) srt_lines.append("") with open(f"{output_dir}/captions_final_v2.srt", "w", encoding="utf-8") as f: f.write("\n".join(srt_lines)) print(f"SRT saved: {len(chunks_with_time)} chunks") # --- Generate ASS --- ass_header = """[Script Info] ; V2 subtitles for full story douyin video ScriptType: v4.00+ PlayResX: 1080 PlayResY: 1920 WrapStyle: 1 ScaledBorderAndShadow: yes [V4+ Styles] Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding Style: Default,Microsoft YaHei,52,&H00FFFFFF,&H0000FF00,&H00000000,&H64000000,0,0,0,0,100,100,0,0,1,3,1,2,60,60,120,1 Style: Highlight,Microsoft YaHei,52,&H0000FFFF,&H0000FF00,&H00000000,&H64000000,0,0,0,0,100,100,0,0,1,3,1,2,60,60,120,1 [Events] Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text """ def has_highlight_word(lines): """Check if any line contains a highlight keyword.""" combined = " ".join(lines) for kw in HIGHLIGHT_KEYWORDS: if kw in combined: return True return False def build_ass_text(lines): """Build ASS text with keyword highlighting.""" combined = "\\N".join(lines) result = combined for kw in sorted(HIGHLIGHT_KEYWORDS, key=len, reverse=True): if kw in result: replacement = r"{\c&H00FFFF&\b1}" + kw + r"{\c&HFFFFFF&\b0}" result = result.replace(kw, replacement) return result ass_events = [] for start, end, lines in chunks_with_time: ass_text = build_ass_text(lines) style = "Highlight" if has_highlight_word(lines) else "Default" ass_events.append( f"Dialogue: 0,{to_ass_time(start)},{to_ass_time(end)},{style},,0,0,0,,{ass_text}" ) ass_content = ass_header + "\n".join(ass_events) with open(f"{output_dir}/captions_final_v2.ass", "w", encoding="utf-8") as f: f.write(ass_content) print(f"ASS saved: {len(ass_events)} events") # --- Generate captions_words_v2.json --- captions_words = { "whisper_model": "base+manual_align", "base_voiceover": "voiceover_master_v2.wav", "total_duration_sec": 91.37, "segments": [] } for start, end, lines in chunks_with_time: combined_text = "".join(lines) chars = list(combined_text) char_duration = (end - start) / max(len(chars), 1) if chars else 0 words = [] for j, ch in enumerate(chars): words.append({ "word": ch, "start": round(start + j * char_duration, 2), "end": round(start + (j + 1) * char_duration, 2) }) captions_words["segments"].append({ "start": start, "end": end, "text": "\n".join(lines), "words": words }) with open(f"{output_dir}/captions_words_v2.json", "w", encoding="utf-8") as f: json.dump(captions_words, f, ensure_ascii=False, indent=2) print(f"captions_words_v2.json updated: {len(chunks_with_time)} segments") print("\nAll subtitle files generated successfully!")