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.
@@ -0,0 +1,5 @@
1
+ """mediakit — agent-friendly video/audio analysis and editing toolkit."""
2
+
3
+ __version__ = "0.1.2"
4
+
5
+ __all__ = ["__version__"]
@@ -0,0 +1,4 @@
1
+ from .cli import main
2
+
3
+ if __name__ == "__main__":
4
+ raise SystemExit(main())
mediakit_cli/auto.py ADDED
@@ -0,0 +1,217 @@
1
+ from __future__ import annotations
2
+
3
+ import json
4
+ import random
5
+ from pathlib import Path
6
+
7
+ from .edit import render
8
+ from .music import analyze
9
+ from .util import ensure_dir, media_path
10
+
11
+ VIDEO_EXT = {".mp4", ".mov", ".mkv", ".webm", ".avi", ".m4v"}
12
+
13
+
14
+ def collect_sources(srcs) -> list:
15
+ files = []
16
+ for raw in srcs:
17
+ p = Path(raw).expanduser()
18
+ if p.is_dir():
19
+ files += [f for f in sorted(p.rglob("*")) if f.suffix.lower() in VIDEO_EXT]
20
+ elif p.suffix.lower() in VIDEO_EXT:
21
+ files.append(media_path(p))
22
+ # 去重且保持顺序
23
+ seen, out = set(), []
24
+ for f in files:
25
+ key = str(f.resolve()).lower()
26
+ if key not in seen:
27
+ seen.add(key)
28
+ out.append(f.resolve())
29
+ return out
30
+
31
+
32
+ def pace_map(sections: list, default_k: int = 4, k_fast: int = 3, k_slow: int = 8) -> list:
33
+ """按段落能量定每段的“几拍一切”:能量越高切得越密。
34
+
35
+ 能量差小于 15% 时视为同一档,避免两段几乎一样响却切出 6 拍 vs 2 拍。
36
+ """
37
+ if not sections or len(sections) < 2:
38
+ return []
39
+ energies = [float(s.get("energy") or 0) for s in sections]
40
+ lo, hi = min(energies), max(energies)
41
+ spread = (hi - lo) / max(hi, 1e-9)
42
+ uniform = spread < 0.15
43
+ out = []
44
+ for s, e in zip(sections, energies):
45
+ if uniform:
46
+ k = default_k
47
+ else:
48
+ norm = (e - lo) / (hi - lo) if hi > lo else 0.5
49
+ k = int(round(k_slow + (k_fast - k_slow) * norm))
50
+ out.append(
51
+ {
52
+ "start": s["start"],
53
+ "end": s["end"],
54
+ "energy": e,
55
+ "beats_per_cut": max(1, min(k, k_slow)),
56
+ }
57
+ )
58
+ return out
59
+
60
+
61
+ def cut_points(
62
+ beats: list,
63
+ duration: float,
64
+ beats_per_cut: int,
65
+ pmap: list | None = None,
66
+ fallback: float = 2.5,
67
+ ) -> list:
68
+ if not beats:
69
+ n = max(1, int(duration // fallback))
70
+ return [round(i * fallback, 3) for i in range(n + 1)] + [round(duration, 3)]
71
+ cuts = [0.0]
72
+ if pmap:
73
+ for sec in pmap:
74
+ k = int(sec["beats_per_cut"])
75
+ bs = [b for b in beats if sec["start"] <= b < sec["end"]]
76
+ for i, b in enumerate(bs):
77
+ if (i + 1) % k == 0 and b > cuts[-1] + 0.4 and b < duration - 0.4:
78
+ cuts.append(round(b, 3))
79
+ else:
80
+ for i, b in enumerate(beats):
81
+ if (i + 1) % max(1, beats_per_cut) == 0 and b > cuts[-1] + 0.4 and b < duration - 0.4:
82
+ cuts.append(round(b, 3))
83
+ cuts.append(round(duration, 3))
84
+ return cuts
85
+
86
+
87
+ def build_pool(files: list, limit: int, min_shot: float, progress) -> list:
88
+ from .shots import detect_shots
89
+
90
+ pool = []
91
+ for idx, f in enumerate(files[:limit], 1):
92
+ try:
93
+ shots = detect_shots(f)
94
+ except Exception as exc:
95
+ progress(f"跳过 {f.name}: {exc}")
96
+ continue
97
+ kept = [s for s in shots if s["dur"] >= min_shot]
98
+ progress(f"[{idx}/{min(len(files), limit)}] {f.name}: {len(shots)} 镜头,{len(kept)} 个可用")
99
+ for s in kept:
100
+ pool.append({"src": str(f), "start": s["start"], "end": s["end"], "dur": s["dur"]})
101
+ return pool
102
+
103
+
104
+ def select_clips(cuts: list, pool: list, seed: int, progress) -> tuple:
105
+ if not pool:
106
+ raise RuntimeError("素材里没有可用镜头:检查 --src 路径或降低 --min-shot")
107
+ rng = random.Random(seed)
108
+ clips, used, last_key = [], set(), None
109
+ clamped, reused = 0, 0
110
+ cursor = 0.0
111
+ for cut in cuts[1:]:
112
+ need = cut - cursor
113
+ if need < 0.4:
114
+ continue
115
+ cands = [s for s in pool if s["dur"] >= need - 1e-3]
116
+ if not cands:
117
+ cands = [max(pool, key=lambda s: s["dur"])]
118
+ clamped += 1
119
+ fresh = [s for s in cands if (s["src"], s["start"]) not in used]
120
+ avoid_last = [s for s in (fresh or cands) if (s["src"], s["start"]) != last_key]
121
+ pick_pool = avoid_last or fresh or cands
122
+ shot = rng.choice(pick_pool)
123
+ if shot["dur"] < need:
124
+ need = shot["dur"]
125
+ key = (shot["src"], shot["start"])
126
+ if key in used:
127
+ reused += 1
128
+ used.add(key)
129
+ last_key = key
130
+ hi = max(shot["start"], shot["end"] - need)
131
+ in_t = rng.uniform(shot["start"], hi)
132
+ clips.append(
133
+ {
134
+ "src": shot["src"],
135
+ "in": round(in_t, 3),
136
+ "out": round(in_t + need, 3),
137
+ }
138
+ )
139
+ cursor += need
140
+ progress(f"选了 {len(clips)} 段,复用 {reused} 次,素材不够被压短 {clamped} 次")
141
+ return clips, {"reused": reused, "clamped": clamped, "filled": round(cursor, 3)}
142
+
143
+
144
+ def generate(
145
+ music,
146
+ srcs,
147
+ out,
148
+ size=(1080, 1920),
149
+ fps: int = 30,
150
+ fit: str = "crop",
151
+ beats_per_cut: int = 4,
152
+ pace: str = "auto",
153
+ transition: str = "none",
154
+ transition_duration: float = 0.25,
155
+ max_sources: int = 12,
156
+ min_shot: float = 0.8,
157
+ seed: int = 7,
158
+ gain_db: float = -14.0,
159
+ plan_only: bool = False,
160
+ progress=None,
161
+ ) -> dict:
162
+ note = progress or (lambda *_: None)
163
+ m_path = media_path(music)
164
+ note(f"分析音乐 {m_path.name}")
165
+ m = analyze(m_path, with_beats=True)
166
+ pmap = pace_map(m.get("sections") or [], default_k=beats_per_cut) if pace == "auto" else []
167
+ cuts = cut_points(m["beats"], m["duration"], beats_per_cut, pmap=pmap)
168
+ if pace == "auto" and len(cuts) < 3:
169
+ cuts = cut_points(m["beats"], m["duration"], beats_per_cut)
170
+ pmap = []
171
+ note("段落变化不足,退回固定拍数")
172
+ note(f"{m['tempo_bpm']} BPM,{len(m['beats'])} 拍,切点 {len(cuts) - 1} 个,节奏 {pace}")
173
+
174
+ files = collect_sources(srcs)
175
+ note(f"素材 {len(files)} 个文件,本次扫描前 {min(len(files), max_sources)} 个")
176
+ pool = build_pool(files, max_sources, min_shot, note)
177
+ clips, stats = select_clips(cuts, pool, seed, note)
178
+
179
+ plan = {
180
+ "size": list(size),
181
+ "fps": fps,
182
+ "fit": fit,
183
+ "crf": 20,
184
+ "clips": clips,
185
+ "music": {"src": str(m_path), "gain_db": gain_db, "loop": True},
186
+ "keep_audio": False,
187
+ }
188
+ if transition and transition != "none":
189
+ plan["transition"] = {"type": transition, "duration": transition_duration}
190
+ out_path = Path(out).resolve()
191
+ ensure_dir(out_path.parent)
192
+ plan_path = out_path.with_suffix(out_path.suffix + ".plan.json")
193
+ plan_path.write_text(json.dumps(plan, ensure_ascii=False, indent=2), encoding="utf-8")
194
+ note(f"计划已写入 {plan_path}")
195
+
196
+ result = {
197
+ "out": None,
198
+ "plan": str(plan_path),
199
+ "duration": m["duration"],
200
+ "tempo_bpm": m["tempo_bpm"],
201
+ "beats": len(m["beats"]),
202
+ "cuts": cuts,
203
+ "segments": len(clips),
204
+ "sources": len(files),
205
+ "pool_shots": len(pool),
206
+ "seed": seed,
207
+ "stats": stats,
208
+ "pace": {"mode": pace, "sections": pmap},
209
+ "transition": transition if transition != "none" else None,
210
+ }
211
+ if plan_only:
212
+ return result
213
+
214
+ note("开始渲染")
215
+ rendered = render(plan, out_path)
216
+ result.update({"out": rendered["out"], "duration": rendered["duration"]})
217
+ return result