mediakit-cli 0.1.2__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
mediakit_cli/cli.py ADDED
@@ -0,0 +1,665 @@
1
+ from __future__ import annotations
2
+
3
+ import argparse
4
+ import json
5
+ import sys
6
+ from pathlib import Path
7
+
8
+ from . import __version__
9
+ from .config import ollama_base, vlm_model
10
+ from .util import ensure_dir, human_duration, out_json, grab_frame
11
+
12
+
13
+ def _cmd_probe(args) -> dict:
14
+ from .probe import probe
15
+
16
+ return probe(args.file)
17
+
18
+
19
+ def _cmd_shots(args) -> dict:
20
+ from .shots import contact_sheet, detect_shots, write_otio
21
+ from .util import media_path
22
+
23
+ src = media_path(args.file)
24
+ shots = detect_shots(src, threshold=args.threshold, min_seconds=args.min_len)
25
+ result = {"path": str(src), "count": len(shots), "shots": shots, "files": {}}
26
+ if args.out:
27
+ out = ensure_dir(args.out)
28
+ (out / "shots.json").write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
29
+ result["files"]["shots"] = str(out / "shots.json")
30
+ if args.sheet:
31
+ sheet = contact_sheet(src, shots, out / "contact.jpg", limit=args.sheet)
32
+ result["files"]["contact"] = str(sheet)
33
+ if args.otio:
34
+ otio = write_otio(src, shots, out / "timeline.otio")
35
+ result["files"]["otio"] = str(otio)
36
+ return result
37
+
38
+
39
+ def _cmd_frames(args) -> dict:
40
+ from .probe import probe
41
+ from .util import media_path
42
+
43
+ src = media_path(args.file)
44
+ info = probe(src)
45
+ duration = float(info.get("duration") or 0)
46
+ out = ensure_dir(args.out)
47
+ n = max(1, args.n)
48
+ items = []
49
+ for i in range(n):
50
+ t = round(duration * (i + 0.5) / n, 3)
51
+ dst = out / f"f{i + 1:03d}.jpg"
52
+ grab_frame(src, t, dst, width=args.width)
53
+ items.append({"t": t, "path": str(dst)})
54
+ return {"path": str(src), "duration": round(duration, 3), "count": n, "frames": items}
55
+
56
+
57
+ def _cmd_subs(args) -> dict:
58
+ from .subs import build_lines, transcribe, write_subtitles
59
+ from .util import media_path
60
+
61
+ src = media_path(args.file)
62
+ asr = transcribe(src, language=args.lang, model=args.model, vad=not args.no_vad)
63
+ lines = build_lines(
64
+ asr, max_chars=args.max_chars, max_dur=args.max_dur, pause=args.pause
65
+ )
66
+ formats = tuple(f.strip() for f in args.formats.split(",") if f.strip())
67
+ res = None
68
+ if args.size:
69
+ try:
70
+ w, h = args.size.lower().replace("*", "x").split("x")
71
+ res = (int(w), int(h))
72
+ except Exception:
73
+ raise ValueError("--size 要写成 1080x1920 这种形式")
74
+ out = ensure_dir(args.out)
75
+ files = write_subtitles(
76
+ lines,
77
+ out,
78
+ stem=args.stem,
79
+ formats=formats,
80
+ font_size=args.font_size,
81
+ resolution=res,
82
+ karaoke=args.karaoke,
83
+ )
84
+ result = {
85
+ "path": str(src),
86
+ "language": asr["language"],
87
+ "duration": asr["duration"],
88
+ "segments": len(asr["segments"]),
89
+ "words": len(asr["words"]),
90
+ "lines": len(lines),
91
+ "files": files,
92
+ "preview": lines[:8],
93
+ "asr_segments": asr["segments"],
94
+ }
95
+ if args.qa:
96
+ from .subs import quality_report
97
+
98
+ result["qa"] = quality_report(lines, max_chars=args.max_chars)
99
+ return result
100
+
101
+
102
+ def _cmd_subqa(args) -> dict:
103
+ import pysubs2
104
+
105
+ from .subs import quality_report
106
+ from .util import media_path
107
+
108
+ subs = pysubs2.load(str(media_path(args.file)))
109
+ lines = [
110
+ {"text": ev.plaintext, "start": ev.start / 1000.0, "end": ev.end / 1000.0}
111
+ for ev in subs
112
+ if ev.plaintext.strip()
113
+ ]
114
+ report = quality_report(
115
+ lines,
116
+ max_cps_cjk=args.max_cps_cjk,
117
+ max_cps_latin=args.max_cps_latin,
118
+ max_chars=args.max_chars,
119
+ )
120
+ report["file"] = str(args.file)
121
+ return report
122
+
123
+
124
+ def _cmd_ocr(args) -> dict:
125
+ import json
126
+
127
+ from .ocr import ocr_video, write_ocr_srt
128
+
129
+ result = ocr_video(
130
+ args.file,
131
+ every=args.every,
132
+ max_frames=args.max_frames,
133
+ min_score=args.min_score,
134
+ progress=lambda m: print(f"[ocr] {m}", file=sys.stderr),
135
+ )
136
+ out = ensure_dir(args.out)
137
+ json_path = out / "ocr.json"
138
+ json_path.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
139
+ result["files"] = {"json": str(json_path)}
140
+ if result["segments"]:
141
+ result["files"]["srt"] = write_ocr_srt(result, out / "ocr.srt")
142
+ return result
143
+
144
+
145
+ def _cmd_index(args) -> dict:
146
+ from .index import build_index
147
+
148
+ return build_index(
149
+ args.file,
150
+ args.out,
151
+ every=args.every,
152
+ max_frames=args.max_frames,
153
+ progress=lambda m: print(f"[index] {m}", file=sys.stderr),
154
+ )
155
+
156
+
157
+ def _cmd_search(args) -> dict:
158
+ from .index import search
159
+
160
+ return search(args.dir, args.query, topk=args.top)
161
+
162
+
163
+ def _cmd_lyrics(args) -> dict:
164
+ from .lyrics import make_lyrics
165
+
166
+ return make_lyrics(
167
+ args.file,
168
+ args.out,
169
+ lang=args.lang,
170
+ model=args.model,
171
+ separate=not args.no_separate,
172
+ max_chars=args.max_chars,
173
+ pause=args.pause,
174
+ karaoke=not args.no_karaoke,
175
+ resolution=_parse_size(args.size, (1920, 1080)),
176
+ sep_model=args.sep_model,
177
+ progress=lambda m: print(f"[lyrics] {m}", file=sys.stderr),
178
+ )
179
+
180
+
181
+ def _cmd_dub(args) -> dict:
182
+ from .dub import dub
183
+
184
+ return dub(
185
+ args.file,
186
+ args.srt,
187
+ args.out,
188
+ voice_name=args.voice,
189
+ rate=args.rate,
190
+ mix_original=args.mix_original,
191
+ fit=not args.no_fit,
192
+ progress=lambda m: print(f"[dub] {m}", file=sys.stderr),
193
+ )
194
+
195
+
196
+ def _cmd_voices(args) -> dict:
197
+ from .dub import list_voices
198
+
199
+ return {"voices": list_voices()}
200
+
201
+
202
+ def _cmd_shotcard(args) -> dict:
203
+ from .shotcard import analyze
204
+
205
+ return analyze(
206
+ args.file,
207
+ args.out,
208
+ max_shots=args.max_shots,
209
+ use_vlm=args.vlm,
210
+ vlm_max=args.vlm_max,
211
+ vlm_model=args.vlm_model,
212
+ progress=lambda m: print(f"[shotcard] {m}", file=sys.stderr),
213
+ )
214
+
215
+
216
+ def _cmd_music(args) -> dict:
217
+ from .music import analyze
218
+ from .util import media_path
219
+
220
+ result = analyze(
221
+ media_path(args.file),
222
+ with_beats=not args.no_beats,
223
+ with_chords=args.chords,
224
+ with_melody=args.melody,
225
+ melody_max=args.melody_max,
226
+ engine=args.engine,
227
+ )
228
+ if args.out:
229
+ out = Path(args.out)
230
+ ensure_dir(out.parent)
231
+ out.write_text(json.dumps(result, ensure_ascii=False, indent=2), encoding="utf-8")
232
+ result["out"] = str(out.resolve())
233
+ return result
234
+
235
+
236
+ def _cmd_watch(args) -> dict:
237
+ from .watch import build_card
238
+
239
+ return build_card(
240
+ args.file,
241
+ args.out,
242
+ lang=args.lang,
243
+ model=args.model,
244
+ max_frames=args.frames,
245
+ do_asr=not args.no_asr,
246
+ do_subs=not args.no_subs,
247
+ do_ocr=args.ocr,
248
+ progress=lambda m: print(f"[watch] {m}", file=sys.stderr),
249
+ )
250
+
251
+
252
+ def _cmd_edit(args) -> dict:
253
+ from .edit import load_plan, render
254
+
255
+ plan = load_plan(args.plan)
256
+ return render(plan, args.out)
257
+
258
+
259
+ def _parse_size(text: str, default=(1080, 1920)):
260
+ if not text:
261
+ return default
262
+ try:
263
+ w, h = text.lower().replace("*", "x").split("x")
264
+ return (int(w), int(h))
265
+ except Exception:
266
+ raise ValueError("尺寸要写成 1080x1920 这种形式")
267
+
268
+
269
+ def _cmd_auto(args) -> dict:
270
+ from .auto import generate
271
+
272
+ return generate(
273
+ args.music,
274
+ args.src,
275
+ args.out,
276
+ size=_parse_size(args.size),
277
+ fps=args.fps,
278
+ fit=args.fit,
279
+ beats_per_cut=args.beats_per_cut,
280
+ pace=args.pace,
281
+ transition=args.transition,
282
+ transition_duration=args.transition_duration,
283
+ max_sources=args.max_sources,
284
+ min_shot=args.min_shot,
285
+ seed=args.seed,
286
+ gain_db=args.gain_db,
287
+ plan_only=args.plan_only,
288
+ progress=lambda m: print(f"[auto] {m}", file=sys.stderr),
289
+ )
290
+
291
+
292
+ def _cmd_polish(args) -> dict:
293
+ from .polish import polish
294
+
295
+ return polish(
296
+ args.file,
297
+ args.out,
298
+ model=args.model,
299
+ endpoint=args.endpoint,
300
+ max_cjk=args.max_cjk,
301
+ split=not args.no_split,
302
+ optimize=not args.no_optimize,
303
+ )
304
+
305
+
306
+ # ---------- human-readable output ----------
307
+
308
+ def _print_probe(info: dict) -> None:
309
+ v = info.get("video") or {}
310
+ a = info.get("audio") or {}
311
+ print(f"文件 {info['path']}")
312
+ print(f"时长 {human_duration(info['duration'])} ({info['duration']}s)")
313
+ print(f"体积 {info['size_bytes'] / 1024 / 1024:.2f} MB 码率 {info['bit_rate'] // 1000} kbps")
314
+ if v:
315
+ print(f"画面 {v['width']}x{v['height']} @ {v['fps']}fps {v['codec']} {v['pix_fmt']}")
316
+ if a:
317
+ print(f"声音 {a['channels']}ch {a['sample_rate']}Hz {a['codec']}")
318
+ if not a:
319
+ print("声音 (无音轨)")
320
+
321
+
322
+ def _print_shots(res: dict) -> None:
323
+ print(f"共 {res['count']} 个镜头")
324
+ for s in res["shots"][:15]:
325
+ print(f" {s['index']:02d} {human_duration(s['start'])} → {human_duration(s['end'])} {s['dur']:.2f}s")
326
+ if res["count"] > 15:
327
+ print(f" … 其余 {res['count'] - 15} 个")
328
+ for k, v in (res.get("files") or {}).items():
329
+ print(f" {k}: {v}")
330
+
331
+
332
+ def _print_subs(res: dict) -> None:
333
+ print(f"语言 {res['language']} 时长 {human_duration(res['duration'])} 词 {res['words']} 个")
334
+ print(f"断成 {res['lines']} 行:")
335
+ for ln in res["preview"]:
336
+ print(f" {ln['start']:7.2f} → {ln['end']:7.2f} {ln['text']}")
337
+ if res["lines"] > len(res["preview"]):
338
+ print(f" … 其余 {res['lines'] - len(res['preview'])} 行")
339
+ for fmt, path in res["files"].items():
340
+ print(f" {fmt}: {path}")
341
+
342
+
343
+ def _print_subqa(res: dict) -> None:
344
+ print(f"{res['file']}")
345
+ print(f"{res['lines']} 条 · 平均 {res['avg_cps']} 字/秒 · 最高 {res['max_cps']} 字/秒")
346
+ if res["ok"]:
347
+ print("没有发现问题")
348
+ return
349
+ print(f"{res['issue_count']} 个问题:")
350
+ for issue in res["issues"][:20]:
351
+ print(f" 第 {issue['line']} 条 @{issue['time']}s [{issue['type']}] {issue['detail']}")
352
+ if res["issue_count"] > 20:
353
+ print(f" … 其余 {res['issue_count'] - 20} 个")
354
+
355
+
356
+ def _print_ocr(res: dict) -> None:
357
+ print(f"{res['path']}")
358
+ print(f"每 {res['every']}s 一帧 · {len(res['frames'])} 帧 · 合并成 {len(res['segments'])} 段")
359
+ for seg in res["segments"][:8]:
360
+ print(f" {seg['start']:7.2f} → {seg['end']:7.2f} {seg['text']}")
361
+ if len(res["segments"]) > 8:
362
+ print(f" … 其余 {len(res['segments']) - 8} 段")
363
+ for k, v in (res.get("files") or {}).items():
364
+ print(f" {k}: {v}")
365
+
366
+
367
+ def _print_index(res: dict) -> None:
368
+ print(f"索引 {res['index']}")
369
+ print(f"{res['frames']} 帧 · 每 {res['every']}s · {res['model']}")
370
+
371
+
372
+ def _print_search(res: dict) -> None:
373
+ print(f"「{res['query']}」在 {res['video']}")
374
+ for hit in res["hits"]:
375
+ print(f" {hit['t']:8.2f}s score {hit['score']:.3f} {hit['thumb']}")
376
+
377
+
378
+ def _print_lyrics(res: dict) -> None:
379
+ print(f"歌曲 {res['song']}")
380
+ if res["vocals"]:
381
+ print(f"人声 {res['vocals']}")
382
+ print(f"语言 {res['language']} · {res['duration']}s · {res['words']} 词 · {res['lines']} 行")
383
+ for ln in res["preview"]:
384
+ print(f" {ln['start']:7.2f} → {ln['end']:7.2f} {ln['text']}")
385
+ for k, v in res["files"].items():
386
+ print(f" {k}: {v}")
387
+
388
+
389
+ def _print_dub(res: dict) -> None:
390
+ print(f"输出 {res['out']}")
391
+ print(f"{res['lines']} 句 · 音色 {res['voice']} · 语速 {res['rate']}")
392
+ if res.get("squeezed"):
393
+ print(f"{res['squeezed']} 句超长,已加速压进字幕时长(最大 {res['max_tempo']}×)")
394
+ if res["mix_original"]:
395
+ print("原声保留 25%,与配音混合")
396
+
397
+
398
+ def _print_voices(res: dict) -> None:
399
+ for v in res["voices"]:
400
+ print(v)
401
+
402
+
403
+ def _print_shotcard(res: dict) -> None:
404
+ s = res["summary"]
405
+ print(f"{res['path']} · {res['count']} 个镜头 · {res['duration']}s")
406
+ print(
407
+ f"平均亮度 {s['avg_brightness']} · 平均饱和 {s['avg_saturation']} · "
408
+ f"静止 {s['static_shots']} / 运动 {s['moving_shots']}"
409
+ )
410
+ for shot in res["shots"][:6]:
411
+ colors = " ".join(c["hex"] for c in shot["colors"][:3])
412
+ print(
413
+ f" {shot['index']:02d} {shot['start']:.2f}-{shot['end']:.2f}s "
414
+ f"亮 {shot['brightness']:.2f} 饱和 {shot['saturation']:.2f} "
415
+ f"{shot['motion']['label']}({shot['motion']['magnitude']}) {colors}"
416
+ )
417
+ if shot.get("vlm") and shot["vlm"].get("caption"):
418
+ v = shot["vlm"]
419
+ print(f" {v.get('shot_size', '?')} · {v.get('camera', '?')} · {v.get('mood', '?')} · {v['caption']}")
420
+ if res["count"] > 6:
421
+ print(f" … 其余 {res['count'] - 6} 个")
422
+
423
+
424
+ def _print_music(res: dict) -> None:
425
+ key = res.get("key") or {}
426
+ print(f"时长 {human_duration(res['duration'])} 速度 {res['tempo_bpm']} BPM 拍数 {res['beat_count']}")
427
+ if res.get("engine") and res["engine"] != "librosa":
428
+ print(f"节拍引擎 {res['engine']} 强拍 {len(res.get('downbeats') or [])} 个")
429
+ print(f"调性 {key.get('label')}(相关度 {key.get('correlation')})")
430
+ print(f"响度 {res['loudness_dbfs']} dBFS 峰值 {res['peak_dbfs']} dBFS")
431
+ print(f"段落 {len(res['sections'])} 段:")
432
+ for sec in res["sections"]:
433
+ print(f" {human_duration(sec['start'])} → {human_duration(sec['end'])} energy {sec['energy']}")
434
+ if res.get("chords"):
435
+ line = " → ".join(c["chord"] for c in res["chords"][:14])
436
+ print(f"和弦 {len(res['chords'])} 段:{line}")
437
+ if res.get("melody"):
438
+ notes = " ".join(m["note"] for m in res["melody"][:10])
439
+ print(f"旋律 {len(res['melody'])} 个音:{notes} …")
440
+
441
+
442
+ def _print_watch(card: dict) -> None:
443
+ info = card["probe"]
444
+ print(f"文件 {info['path']}")
445
+ print(f"时长 {human_duration(info['duration'])} 镜头 {len(card['shots'])} 个 关键帧 {len(card['keyframes'])} 张")
446
+ print(f"转写 {len(card['transcript'])} 段" + (f" (失败:{card['asr_error']})" if card["asr_error"] else ""))
447
+ for seg in card["transcript"][:6]:
448
+ print(f" {seg['start']:7.2f} → {seg['end']:7.2f} {seg['text']}")
449
+ if len(card["transcript"]) > 6:
450
+ print(f" … 其余 {len(card['transcript']) - 6} 段")
451
+ for k, v in card["files"].items():
452
+ if isinstance(v, dict):
453
+ for kk, vv in v.items():
454
+ print(f" {kk}: {vv}")
455
+ elif v:
456
+ print(f" {k}: {v}")
457
+
458
+
459
+ def _print_auto(res: dict) -> None:
460
+ print(f"输出 {res.get('out') or '(只出计划,未渲染)'}")
461
+ print(
462
+ f"{res['tempo_bpm']} BPM · {res['segments']} 段 · "
463
+ f"{res['pool_shots']} 个候选镜头 / {res['sources']} 个素材文件"
464
+ )
465
+ print(f"种子 {res['seed']} · 复用 {res['stats']['reused']} 次 · 素材不够压短 {res['stats']['clamped']} 次")
466
+ print(f"计划 {res['plan']}")
467
+
468
+
469
+ def _print_polish(res: dict) -> None:
470
+ print(f"输入 {res['in']}")
471
+ print(f"输出 {res['out']} ({res['lines']} 行,{res['bytes']} 字节)")
472
+ print(f"模型 {res['model']} @ {res['endpoint']}")
473
+ print("注意:VideoCaptioner 优化完会主动去标点;4B 级模型偶有改写,先小样对比再全量")
474
+
475
+
476
+ def _print_edit(res: dict) -> None:
477
+ print(f"输出 {res['out']}")
478
+ print(f"{res['clips']} 个片段 {res['duration']:.2f}s {res['size'][0]}x{res['size'][1]} @ {res['fps']}fps")
479
+ print(f"字幕 {'有' if res['subtitled'] else '无'} 配乐 {'有' if res['music'] else '无'} 原声 {'保留' if res['keep_audio'] else '丢弃'}")
480
+ for note in res.get("notes") or []:
481
+ print(f" 注意: {note}")
482
+
483
+
484
+ def build_parser() -> argparse.ArgumentParser:
485
+ parser = argparse.ArgumentParser(
486
+ prog="mk", description="mediakit:把看视频/听音乐/做字幕/剪片子合成一套 CLI(供人或 agent 调用)"
487
+ )
488
+ parser.add_argument("--version", action="version", version=f"mediakit {__version__}")
489
+ parser.add_argument("--json", action="store_true", help="输出机器可读 JSON")
490
+ sub = parser.add_subparsers(dest="cmd", required=True)
491
+
492
+ p = sub.add_parser("probe", help="读媒体元信息")
493
+ p.add_argument("file")
494
+ p.set_defaults(func=_cmd_probe, printer=_print_probe)
495
+
496
+ p = sub.add_parser("shots", help="切镜头 + 联系表 + OTIO 时间轴")
497
+ p.add_argument("file")
498
+ p.add_argument("--out", help="输出目录")
499
+ p.add_argument("--threshold", type=float, default=27.0)
500
+ p.add_argument("--min-len", type=float, default=0.5, help="最短镜头秒数")
501
+ p.add_argument("--sheet", type=int, default=9, help="联系表最多几张,0 表示不生成")
502
+ p.add_argument("--otio", action="store_true", help="导出 timeline.otio")
503
+ p.set_defaults(func=_cmd_shots, printer=_print_shots)
504
+
505
+ p = sub.add_parser("frames", help="均匀抽帧")
506
+ p.add_argument("file")
507
+ p.add_argument("--n", type=int, default=12)
508
+ p.add_argument("--out", required=True)
509
+ p.add_argument("--width", type=int, default=960)
510
+ p.set_defaults(func=_cmd_frames, printer=lambda r: print(f"抽出 {r['count']} 张 → {r['frames'][0]['path'].rsplit(chr(92), 1)[0]}"))
511
+
512
+ p = sub.add_parser("subs", help="转写 + 断句,出 SRT/ASS")
513
+ p.add_argument("file")
514
+ p.add_argument("--lang", default=None, help="zh / en / ja …,不填自动识别")
515
+ p.add_argument("--model", default="small", help="tiny/base/small/medium/large-v3")
516
+ p.add_argument("--max-chars", type=float, default=16.0, help="每行最大显示宽度(汉字算 1)")
517
+ p.add_argument("--max-dur", type=float, default=6.0)
518
+ p.add_argument("--pause", type=float, default=0.28, help="停顿多少秒就断行")
519
+ p.add_argument("--formats", default="srt,ass")
520
+ p.add_argument("--stem", default="subs")
521
+ p.add_argument("--font-size", type=int, default=None, help="不填按画面高度自动算")
522
+ p.add_argument("--size", default=None, help="成片尺寸 WxH,例如 1080x1920;决定 ASS 的 PlayRes")
523
+ p.add_argument("--out", default=".")
524
+ p.add_argument("--no-vad", action="store_true")
525
+ p.add_argument("--karaoke", action="store_true", help="ASS 里生成逐字 \kf 卡拉OK")
526
+ p.add_argument("--qa", action="store_true", help="顺带跑一遍阅读速度/行长检查")
527
+ p.set_defaults(func=_cmd_subs, printer=_print_subs)
528
+
529
+ p = sub.add_parser("subqa", help="检查字幕阅读速度/行长/重叠")
530
+ p.add_argument("file")
531
+ p.add_argument("--max-cps-cjk", type=float, default=9.0)
532
+ p.add_argument("--max-cps-latin", type=float, default=20.0)
533
+ p.add_argument("--max-chars", type=float, default=16.0)
534
+ p.set_defaults(func=_cmd_subqa, printer=_print_subqa)
535
+
536
+ p = sub.add_parser("ocr", help="读画面上的字(硬字幕/歌词/录屏)")
537
+ p.add_argument("file")
538
+ p.add_argument("--out", required=True, help="输出目录")
539
+ p.add_argument("--every", type=float, default=2.0, help="每几秒取一帧")
540
+ p.add_argument("--max-frames", type=int, default=60)
541
+ p.add_argument("--min-score", type=float, default=0.5)
542
+ p.set_defaults(func=_cmd_ocr, printer=_print_ocr)
543
+
544
+ p = sub.add_parser("shotcard", help="每个镜头的亮度/饱和/主色/运动")
545
+ p.add_argument("file")
546
+ p.add_argument("--out", required=True)
547
+ p.add_argument("--max-shots", type=int, default=None)
548
+ p.add_argument("--vlm", action="store_true", help="用本机 VLM 给每个镜头写景别/机位/描述")
549
+ p.add_argument("--vlm-max", type=int, default=12, help="最多给几个镜头跑 VLM")
550
+ p.add_argument("--vlm-model", default=None)
551
+ p.set_defaults(func=_cmd_shotcard, printer=_print_shotcard)
552
+
553
+ p = sub.add_parser("index", help="给长视频建 CLIP 索引(可搜片段)")
554
+ p.add_argument("file")
555
+ p.add_argument("--out", required=True)
556
+ p.add_argument("--every", type=float, default=2.0)
557
+ p.add_argument("--max-frames", type=int, default=150)
558
+ p.set_defaults(func=_cmd_index, printer=_print_index)
559
+
560
+ p = sub.add_parser("search", help="用一句话在索引里找片段")
561
+ p.add_argument("dir")
562
+ p.add_argument("query")
563
+ p.add_argument("--top", type=int, default=5)
564
+ p.set_defaults(func=_cmd_search, printer=_print_search)
565
+
566
+ p = sub.add_parser("dub", help="用 Windows 本地语音给字幕配音")
567
+ p.add_argument("file")
568
+ p.add_argument("--srt", required=True)
569
+ p.add_argument("--out", required=True)
570
+ p.add_argument("--voice", default=None, help="音色名的一部分,例如 Huihui")
571
+ p.add_argument("--rate", type=int, default=0, help="-10..10")
572
+ p.add_argument("--mix-original", action="store_true")
573
+ p.add_argument("--no-fit", action="store_true", help="不把过长的配音加速压进字幕时长")
574
+ p.set_defaults(func=_cmd_dub, printer=_print_dub)
575
+
576
+ p = sub.add_parser("voices", help="列出本机 SAPI 语音")
577
+ p.set_defaults(func=_cmd_voices, printer=_print_voices)
578
+
579
+ p = sub.add_parser("lyrics", help="分人声 + 识别歌词,出 SRT/ASS/LRC")
580
+ p.add_argument("file")
581
+ p.add_argument("--out", required=True)
582
+ p.add_argument("--lang", default=None)
583
+ p.add_argument("--model", default="small")
584
+ p.add_argument("--no-separate", action="store_true")
585
+ p.add_argument("--sep-model", default=None)
586
+ p.add_argument("--max-chars", type=float, default=18.0)
587
+ p.add_argument("--pause", type=float, default=0.35)
588
+ p.add_argument("--no-karaoke", action="store_true")
589
+ p.add_argument("--size", default="1920x1080")
590
+ p.set_defaults(func=_cmd_lyrics, printer=_print_lyrics)
591
+
592
+ p = sub.add_parser("music", help="测 BPM/调性/段落")
593
+ p.add_argument("file")
594
+ p.add_argument("--out", help="把结果写成 JSON")
595
+ p.add_argument("--no-beats", action="store_true")
596
+ p.add_argument("--chords", action="store_true", help="估和弦进行(模板匹配)")
597
+ p.add_argument("--melody", action="store_true", help="估主旋律音高(pyin)")
598
+ p.add_argument("--melody-max", type=float, default=40.0, help="旋律只分析前 N 秒")
599
+ p.add_argument("--engine", choices=("librosa", "beat-this"), default="librosa",
600
+ help="节拍追踪器;beat-this 会同时给 downbeats")
601
+ p.set_defaults(func=_cmd_music, printer=_print_music)
602
+
603
+ p = sub.add_parser("watch", help="一张卡片看完一个视频")
604
+ p.add_argument("file")
605
+ p.add_argument("--out", required=True)
606
+ p.add_argument("--lang", default=None)
607
+ p.add_argument("--model", default="small")
608
+ p.add_argument("--frames", type=int, default=12)
609
+ p.add_argument("--no-asr", action="store_true")
610
+ p.add_argument("--no-subs", action="store_true")
611
+ p.add_argument("--ocr", action="store_true", help="顺带读关键帧上的字")
612
+ p.set_defaults(func=_cmd_watch, printer=_print_watch)
613
+
614
+ p = sub.add_parser("auto", help="音乐 + 素材 → 卡点短片")
615
+ p.add_argument("--music", required=True)
616
+ p.add_argument("--src", action="append", required=True, help="素材文件或目录,可重复给")
617
+ p.add_argument("--out", required=True)
618
+ p.add_argument("--size", default="1080x1920")
619
+ p.add_argument("--fps", type=int, default=30)
620
+ p.add_argument("--fit", default="crop", choices=["crop", "pad"])
621
+ p.add_argument("--beats-per-cut", type=int, default=4, help="每几拍切一次(--pace fixed 时生效)")
622
+ p.add_argument("--pace", default="auto", choices=["auto", "fixed"], help="auto=按段落能量变节奏")
623
+ p.add_argument("--max-sources", type=int, default=12)
624
+ p.add_argument("--min-shot", type=float, default=0.8)
625
+ p.add_argument("--seed", type=int, default=7)
626
+ p.add_argument("--gain-db", type=float, default=-14)
627
+ p.add_argument("--plan-only", action="store_true")
628
+ p.add_argument("--transition", default="none", choices=["none", "fade", "dissolve", "wipeleft", "slideleft"])
629
+ p.add_argument("--transition-duration", type=float, default=0.25)
630
+ p.set_defaults(func=_cmd_auto, printer=_print_auto)
631
+
632
+ p = sub.add_parser("polish", help="用本地 Ollama(经 VideoCaptioner)润色字幕")
633
+ p.add_argument("file")
634
+ p.add_argument("--out", required=True)
635
+ p.add_argument("--model", default=None)
636
+ p.add_argument("--endpoint", default=None)
637
+ p.add_argument("--max-cjk", type=int, default=16)
638
+ p.add_argument("--no-split", action="store_true")
639
+ p.add_argument("--no-optimize", action="store_true")
640
+ p.set_defaults(func=_cmd_polish, printer=_print_polish)
641
+
642
+ p = sub.add_parser("edit", help="按 JSON 计划剪一条片子")
643
+ p.add_argument("plan")
644
+ p.add_argument("--out", required=True)
645
+ p.set_defaults(func=_cmd_edit, printer=_print_edit)
646
+
647
+ return parser
648
+
649
+
650
+ def main(argv=None) -> int:
651
+ parser = build_parser()
652
+ args = parser.parse_args(argv)
653
+ try:
654
+ result = args.func(args)
655
+ except Exception as exc:
656
+ sys.stderr.write(f"错误: {exc}\n")
657
+ return 2
658
+ if args.json:
659
+ out_json(result)
660
+ else:
661
+ try:
662
+ args.printer(result)
663
+ except Exception:
664
+ out_json(result)
665
+ return 0