Import video pipeline as it was

1 files changed, 46 insertions, 0 deletions

diff --git a/scripts/regen.py b/scripts/regen.py
new file mode 100644
index 0000000..7ff69a8
--- /dev/null
+++ b/scripts/regen.py
@@ -0,0 +1,46 @@
+# -*- coding: utf-8 -*-
+"""Regenera segmentos puntuales probando varias semillas y quedandose con la mejor."""
+import json, os, subprocess, sys, shutil
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+from narration import SCRIPTS
+from gen_audio import synth_seguro, transcribe, dur, ROOT
+from verify import norm
+from difflib import SequenceMatcher
+
+TRIES = 5
+
+
+def main(name, idxs):
+    outdir = os.path.join(ROOT, "audio", name)
+    work = os.path.join(ROOT, "tmp", name + "_rg")
+    os.makedirs(work, exist_ok=True)
+    data = json.load(open(os.path.join(outdir, "segments.json"), encoding="utf-8"))
+    segs = SCRIPTS[name]["segments"]
+
+    for i in idxs:
+        text = segs[i][0]
+        cands = []
+        for k in range(TRIES):
+            c = os.path.join(work, f"{i:02d}_{k}.wav")
+            synth_seguro(text, c, 4242 + 977 * k, f"[{name}] {i:02d}")
+            cands.append(c)
+        tr = transcribe(cands, work)
+        scored = sorted(((SequenceMatcher(None, norm(text, False),
+                                          norm(tr[c], False)).ratio(), c) for c in cands),
+                        reverse=True)
+        for r, c in scored:
+            print(f"  {i:02d} seed{cands.index(c)} r={r:.2f} {dur(c):.2f}s | {tr[c][:70]}")
+        best_r, best = scored[0]
+        shutil.copy(best, os.path.join(outdir, f"{i:02d}.wav"))
+        data["segments"][i].update(tts=text, sub=segs[i][1],
+                                   dur=round(dur(best), 3), score=round(best_r, 3))
+        print(f"  -> {i:02d} elegido r={best_r:.2f}\n")
+
+    data["total_audio"] = round(sum(s["dur"] for s in data["segments"]), 2)
+    json.dump(data, open(os.path.join(outdir, "segments.json"), "w", encoding="utf-8"),
+              ensure_ascii=False, indent=2)
+    print(f"[{name}] actualizado, audio={data['total_audio']}s")
+
+
+if __name__ == "__main__":
+    main(sys.argv[1], [int(x) for x in sys.argv[2:]])