Build latest book artifacts / build (push) Canceled after 0s
dependency resolution / resolve (3.11) (push) Canceled after 0s
dependency resolution / resolve (3.13) (push) Canceled after 0s
deploy-pages / build (push) Canceled after 0s
deploy-pages / deploy (push) Canceled after 0s
i18n consistency check / check (push) Canceled after 0s
provider adoption tests / test (chapter2/context-compression) (push) Canceled after 0s
provider adoption tests / test (chapter2/prompt-injection) (push) Canceled after 0s
provider adoption tests / test (chapter2/system-hint) (push) Canceled after 0s
provider adoption tests / test (chapter3/log-sanitization) (push) Canceled after 0s
web-search-agent tests / test (push) Canceled after 0s
web-search-agent tests / agentbook (push) Canceled after 0s
127 lines
5.5 KiB
Python
127 lines
5.5 KiB
Python
"""
|
||
视频剪辑执行层(双后端)。
|
||
|
||
书中实验 5-6 的核心是"代码生成":Proposer Agent 生成一段调用 Blender Python API
|
||
(bpy)的脚本来完成剪辑。因此本层有两个后端:
|
||
|
||
- blender:把剪辑计划翻译成 bpy 脚本,用 `blender --background --python` 无头渲染
|
||
(见 blender_editor.py)——书中原方案,需安装 Blender;
|
||
- ffmpeg :用 ffmpeg 完成等价的裁剪/字幕/慢动作,单二进制、CI 友好,本机已验证。
|
||
|
||
apply_edit(..., backend=) 统一入口:backend="auto" 时装了 Blender 走 bpy,否则回退
|
||
ffmpeg。**无论走哪个后端,都会把 Proposer 生成的 bpy 脚本落盘(output/edit.py)**,
|
||
体现"生成 Blender Python API 代码"这一核心,便于人工核对或换机执行。
|
||
|
||
支持的操作:
|
||
- trim 裁剪 [start, end] 片段
|
||
- subtitle 在片段上叠加字幕
|
||
- slowmo 慢动作(放慢到 factor 倍时长)
|
||
所有操作最终产出一个标准 mp4(H.264 + AAC)。
|
||
"""
|
||
import os
|
||
|
||
from blender_editor import blender_available, render_with_blender, write_bpy_script
|
||
from ffmpeg_utils import find_font, run
|
||
|
||
|
||
def _esc(text: str) -> str:
|
||
"""转义 drawtext/subtitles 文本中的特殊字符。"""
|
||
return text.replace("\\", "\\\\").replace(":", r"\:").replace("'", r"\'")
|
||
|
||
|
||
def apply_edit(source: str, plan: dict, out_path: str,
|
||
backend: str = "auto", script_path: str = None) -> str:
|
||
"""
|
||
按剪辑计划 plan 生成成片。
|
||
|
||
plan 结构(由 Proposer Agent 产出):
|
||
{
|
||
"start": float, # 目标片段起点(秒)
|
||
"end": float, # 目标片段终点(秒)
|
||
"effects": [ # 可选特效列表
|
||
{"type": "subtitle", "text": "..."},
|
||
{"type": "slowmo", "factor": 2.0}
|
||
]
|
||
}
|
||
|
||
backend:
|
||
"auto" 装了 blender 用 bpy,否则回退 ffmpeg(默认)。
|
||
"blender" 强制用 Blender bpy 无头渲染(未装 blender 则报错)。
|
||
"ffmpeg" 强制用 ffmpeg 剪辑。
|
||
|
||
无论哪个后端,都会把 Proposer 生成的 bpy 脚本写到 script_path
|
||
(默认 out 同目录 edit.py),作为"生成 Blender API 代码"的产物。
|
||
返回成片路径。
|
||
"""
|
||
os.makedirs(os.path.dirname(out_path) or ".", exist_ok=True)
|
||
if script_path is None:
|
||
script_path = os.path.join(os.path.dirname(out_path) or ".", "edit.py")
|
||
|
||
# 始终产出 bpy 脚本:这正是书中"Proposer 生成 Blender 脚本"的落地产物。
|
||
write_bpy_script(source, plan, out_path, script_path)
|
||
|
||
use_blender = backend == "blender" or (backend == "auto" and blender_available())
|
||
if use_blender:
|
||
return render_with_blender(source, plan, out_path, script_path)
|
||
if backend == "blender": # 显式要求 Blender 却没装
|
||
return render_with_blender(source, plan, out_path, script_path) # 抛清晰错误
|
||
return _apply_edit_ffmpeg(source, plan, out_path)
|
||
|
||
|
||
def _apply_edit_ffmpeg(source: str, plan: dict, out_path: str) -> str:
|
||
"""ffmpeg 后端:与 bpy 脚本等价的裁剪/字幕/慢动作,产出 H.264+AAC 的 mp4。"""
|
||
start, end = float(plan["start"]), float(plan["end"])
|
||
if end <= start:
|
||
raise ValueError(f"剪辑区间非法:start={start} >= end={end}")
|
||
|
||
effects = plan.get("effects", []) or []
|
||
vf_chain = [] # 视频滤镜链
|
||
af_chain = [] # 音频滤镜链
|
||
font = find_font()
|
||
|
||
for eff in effects:
|
||
etype = eff.get("type")
|
||
if etype == "subtitle":
|
||
txt = _esc(eff.get("text", ""))
|
||
opts = [f"text='{txt}'", "fontsize=52", "fontcolor=white",
|
||
"x=(w-text_w)/2", "y=h-text_h-50",
|
||
"box=1", "boxcolor=black@0.6", "boxborderw=16"]
|
||
if font:
|
||
opts.insert(0, f"fontfile={font}")
|
||
vf_chain.append("drawtext=" + ":".join(opts))
|
||
elif etype == "slowmo":
|
||
raw = eff.get("factor", 2.0)
|
||
if raw is None:
|
||
continue
|
||
factor = float(raw)
|
||
if factor <= 0:
|
||
continue
|
||
vf_chain.append(f"setpts={factor}*PTS")
|
||
# atempo 只支持 0.5~2.0,用 1/factor 放慢音频。
|
||
af_chain.append(f"atempo={max(0.5, min(2.0, 1.0 / factor))}")
|
||
|
||
cmd = ["ffmpeg", "-y", "-ss", f"{start:.3f}", "-to", f"{end:.3f}", "-i", source]
|
||
if vf_chain:
|
||
cmd += ["-vf", ",".join(vf_chain)]
|
||
if af_chain:
|
||
cmd += ["-af", ",".join(af_chain)]
|
||
cmd += ["-c:v", "libx264", "-c:a", "aac", "-pix_fmt", "yuv420p", out_path]
|
||
|
||
out_dir = os.path.dirname(out_path)
|
||
if out_dir:
|
||
os.makedirs(out_dir, exist_ok=True)
|
||
run(cmd, desc="ffmpeg 剪辑")
|
||
return out_path
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# Blender 后端说明
|
||
# --------------------------------------------------------------------------- #
|
||
# 真实的 Blender bpy 脚本生成与无头渲染已实现在 blender_editor.py:
|
||
# - generate_bpy_script():把剪辑计划翻译成一段 bpy 脚本(new_movie / 裁剪 /
|
||
# TEXT / SPEED / render)——即书中"Proposer 生成 Blender 脚本";
|
||
# - render_with_blender():用 `blender --background --python edit.py` 无头执行。
|
||
# apply_edit(backend="blender") 即走这条路径;backend="auto"(默认)在未装 Blender 时
|
||
# 回退到本文件的 ffmpeg 后端。核心的"两步 Vision 定位 + 提议者-审核者"与执行层解耦,
|
||
# 两个后端共用同一份剪辑计划(plan),agents.py / demo.py 无需为切换后端改动逻辑。
|