feat: add config-driven extraction and audio sync tooling

extract.py

@@ -12,16 +12,53 @@ import os
 import sys
 import shutil
 
-# ─── Hardcoded input (relative to this script's directory) ───
-inputs = [
-    "/home/honney/Videos/Drachenzähmen/Drachenzaehmen.mp4",
-    "/home/honney/Videos/Drachenzähmen/How to Train Your Dragon.mkv"
-]
+# ─── Input / config (config.toml is base config, fallback to hardcoded) ───
+CONFIG_PATH = os.path.join(os.path.dirname(os.path.abspath(__file__)), "config.toml")
+_cfg_inputs = None
+_cfg_duration = 90
+_cfg_start = 0.0
+_cfg_out_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "extracted")
+_cfg_target_fps = "24000/1001"
+_cfg_eps = 0.02
+try:
+    try:
+        import tomllib  # py3.11+
+    except ModuleNotFoundError:
+        import tomli as tomllib  # type: ignore
+    if os.path.exists(CONFIG_PATH):
+        with open(CONFIG_PATH, "rb") as _f:
+            _cfg = tomllib.load(_f)
+        _ex = _cfg.get("extract", {})
+        if _ex.get("file_a") or _ex.get("file_b"):
+            _cfg_inputs = [p for p in [_ex.get("file_a"), _ex.get("file_b")] if p]
+        _cfg_duration = int(_ex.get("duration", _cfg_duration))
+        _cfg_start = float(_ex.get("start", _cfg_start))
+        _cfg_out_dir = _ex.get("out_dir", _cfg_out_dir)
+        if _ex.get("target_fps"):
+            _cfg_target_fps = str(_ex["target_fps"])
+        if _ex.get("target_fps_eps") is not None:
+            try:
+                _cfg_eps = float(_ex["target_fps_eps"])
+            except Exception:
+                pass
+        if _cfg_inputs:
+            print(f"[config] loaded {CONFIG_PATH}: duration={_cfg_duration}s start={_cfg_start}s target_fps={_cfg_target_fps}")
+except Exception as _e:
+    print(f"[config] warn: failed to load {CONFIG_PATH}: {_e}", file=sys.stderr)
+
+if _cfg_inputs is not None:
+    inputs = _cfg_inputs
+else:
+    inputs = [
+        "/home/honney/Videos/Drachenzähmen/Drachenzaehmen.mp4",
+        "/home/honney/Videos/Drachenzähmen/How to Train Your Dragon.mkv"
+    ]
 INPUT_FILES = []
 for input_file in inputs:
     INPUT_FILES.append(os.path.join(os.path.dirname(os.path.abspath(__file__)), input_file))
-DURATION   = 90  # 1.5 minutes in seconds
-OUT_DIR    = os.path.join(os.path.dirname(os.path.abspath(__file__)), "extracted")
+DURATION   = _cfg_duration
+START_SEC  = _cfg_start  # for ffmpeg -ss (0 = file start)
+OUT_DIR    = os.path.join(os.path.dirname(os.path.abspath(__file__)), _cfg_out_dir) if not os.path.isabs(_cfg_out_dir) else _cfg_out_dir
 
 # ─── Analysis-friendly codecs (NOT GPU) ───
 # Video: FFV1 (lossless, intra-only) in MKV. Every frame stands alone, no
@@ -39,9 +76,14 @@ OUT_DIR    = os.path.join(os.path.dirname(os.path.abspath(__file__)), "extracted
 #   of 90 s clips is fast enough and bit-exact.
 VIDEO_CODEC_BASE = ["-c:v", "ffv1", "-level", "3", "-pix_fmt", "yuv444p"]
 AUDIO_CODEC = ["-c:a", "pcm_f32le", "-ar", "48000", "-ac", "1"]
-TARGET_FPS = "24000/1001"            # Blu-ray film = 23.976023... (shown as 23.98)
-TARGET_FPS_FLOAT = 24000 / 1001
-FPS_EPS = 0.02                        # tolerance for float comparison
+# target fps may be overridden via config.toml
+TARGET_FPS = _cfg_target_fps  # default "24000/1001"
+try:
+    _n, _d = TARGET_FPS.split("/") if "/" in TARGET_FPS else (TARGET_FPS, "1")
+    TARGET_FPS_FLOAT = float(_n) / float(_d)
+except Exception:
+    TARGET_FPS_FLOAT = 24000 / 1001
+FPS_EPS = _cfg_eps  # tolerance for float comparison
 
 def _parse_rate(s: str) -> float | None:
     s = s.strip()
@@ -135,9 +177,11 @@ for INPUT_FILE in INPUT_FILES:
         if not needs_fps:
             video_codec = VIDEO_CODEC_BASE + ["-vf", f"fps={TARGET_FPS}:round=near", "-fps_mode", "cfr"]
 
+    ss_args = ["-ss", str(START_SEC)] if START_SEC and START_SEC != 0 else []
     cmd_video = [
         "ffmpeg",
         "-y",               # overwrite output
+        *ss_args,           # seek before -i (fast, frame-accurate with re-encode)
         "-i", INPUT_FILE,
         "-t", str(DURATION),
         "-an",              # no audio
@@ -154,6 +198,7 @@ for INPUT_FILE in INPUT_FILES:
     cmd_audio = [
         "ffmpeg",
         "-y",
+        *ss_args,
         "-i", INPUT_FILE,
         "-t", str(DURATION),
         "-vn",              # no video