cang-ying/video-ai-system/lib/composer.py

# LIB · 成片合成模块
# D180 · 2026-07-10
"""FFmpeg: 多镜头拼接 + 音频合成 + 字幕叠加 → 导出成片"""
import os, subprocess, json
from lib.models import get_db

FFMPEG = os.path.expanduser("~/.workbuddy/binaries/ffmpeg.exe")
PROJECT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))

def concat_shots(episode_id, output_path=None, add_fade=True):
    """
    从数据库读取已完成镜头 → FFmpeg 拼接 → 导出成片
    
    Args:
        episode_id: 剧集ID
        output_path: 输出路径（默认 outputs/episodes/EP01-final.mp4）
        add_fade: 镜头间加淡入淡出
    """
    if not os.path.exists(FFMPEG):
        return {"error": f"ffmpeg not found: {FFMPEG}"}
    
    db = get_db()
    shots = db.execute("""
        SELECT id, shot_number, output_path, duration_sec FROM shots 
        WHERE episode_id = ? AND status = 'done' 
        ORDER BY shot_number
    """, (episode_id,)).fetchall()
    
    if not shots:
        return {"error": f"No completed shots for {episode_id}"}
    
    # 检查所有文件存在
    for s in shots:
        if not s["output_path"] or not os.path.exists(os.path.join(PROJECT, s["output_path"])):
            return {"error": f"Missing: {s['shot_number']} → {s['output_path']}"}
    
    if not output_path:
        out_dir = os.path.join(PROJECT, "outputs", "episodes")
        os.makedirs(out_dir, exist_ok=True)
        output_path = os.path.join(out_dir, f"{episode_id}-final.mp4")
    else:
        os.makedirs(os.path.dirname(output_path), exist_ok=True)
    
    # 生成 ffmpeg concat 文件列表
    concat_list = os.path.join(PROJECT, "outputs", "temp_concat.txt")
    with open(concat_list, "w", encoding="utf-8") as f:
        for s in shots:
            abs_path = os.path.join(PROJECT, s["output_path"]).replace("\\", "/")
            f.write(f"file '{abs_path}'\n")
            f.write(f"duration {s['duration_sec']}\n")
        # 最后一个文件需要重复一次（ffmpeg concat 约定）
        last = os.path.join(PROJECT, shots[-1]["output_path"]).replace("\\", "/")
        f.write(f"file '{last}'\n")
    
    # FFmpeg concat
    cmd = [
        FFMPEG, "-y",
        "-f", "concat",
        "-safe", "0",
        "-i", concat_list,
        "-c:v", "libx264",
        "-preset", "fast",
        "-crf", "23",
        "-c:a", "aac",
        "-pix_fmt", "yuv420p",
        output_path
    ]
    
    r = subprocess.run(cmd, capture_output=True, text=True)
    os.unlink(concat_list)
    
    if r.returncode != 0:
        return {"error": r.stderr[:500]}
    
    return {
        "output_path": output_path,
        "shot_count": len(shots),
        "total_duration": sum(s["duration_sec"] for s in shots),
        "shots": [s["shot_number"] for s in shots]
    }

def add_audio_and_subtitle(video_path, audio_paths, subtitle_texts, output_path):
    """
    视频 + 配音 + 字幕 → 合成
    
    audio_paths: [{"path": "...", "start": 3.5}, ...]  音频文件和起始时间(秒)
    subtitle_texts: [{"text": "...", "start": 3.5, "end": 8.5}, ...]
    """
    if not os.path.exists(FFMPEG):
        return {"error": f"ffmpeg not found: {FFMPEG}"}
    
    asrc = ""  # no original audio
    
    # Build audio inputs
    audio_inputs = []
    audio_filters = []
    for i, a in enumerate(audio_paths):
        audio_inputs.extend(["-i", a["path"]])
        delay_ms = int(a["start"] * 1000)
        audio_filters.append(f"[{i+1}:a]adelay={delay_ms}|{delay_ms}[a{i}]")
    
    # Build subtitle filter
    sub_parts = []
    for s in subtitle_texts:
        esc = s["text"].replace("'", "\\'").replace(":", "\\:")
        sub_parts.append(f"drawtext=text='{esc}':fontsize=28:fontcolor=white:borderw=2:bordercolor=black:x=(w-text_w)/2:y=h-120:enable='between(t,{s['start']},{s['end']})'")
    sub_filter = ",".join(sub_parts) if sub_parts else "null"
    
    # Audio mix
    if audio_inputs:
        n = len(audio_inputs)
        amix = "".join(audio_filters) + f"amix=inputs={n}:duration=first[aout]"
    else:
        amix = ""
    
    filter_complex = f"[0:v]{sub_filter}[vout];{amix}" if audio_inputs else f"[0:v]{sub_filter}[vout]"
    
    cmd = [FFMPEG, "-y", "-i", video_path] + audio_inputs + [
        "-filter_complex", filter_complex,
        "-map", "[vout]",
        "-map", "[aout]" if audio_inputs else "-an",
        "-c:v", "libx264", "-preset", "fast", "-crf", "23",
        "-c:a", "aac",
        "-pix_fmt", "yuv420p",
        output_path
    ]
    
    r = subprocess.run(cmd, capture_output=True, text=True)
    return output_path if r.returncode == 0 else {"error": r.stderr[:500]}

def render_episode(episode_id):
    """
    一键成片：数据库 → 拼接所有已完成镜头 → 导出
    """
    print(f"🎬 合成 {episode_id}...")
    
    # Step 1: Concat all done shots
    result = concat_shots(episode_id)
    if "error" in result:
        return result
    
    print(f"✅ 拼接完成: {result['shot_count']}镜 {result['total_duration']}s")
    print(f"📁 {result['output_path']}")
    return result

if __name__ == "__main__":
    import sys
    if len(sys.argv) < 2:
        print("Usage: python composer.py concat <episode_id>")
        print("       python composer.py render <episode_id>")
        sys.exit(1)
    cmd = sys.argv[1]
    ep = sys.argv[2] if len(sys.argv) > 2 else "DSV-EP01"
    if cmd == "concat":
        print(json.dumps(concat_shots(ep), indent=2, ensure_ascii=False))
    elif cmd == "render":
        print(json.dumps(render_episode(ep), indent=2, ensure_ascii=False))