1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
|
# -*- coding: utf-8 -*-
"""Regenerates specific segments trying several seeds and keeping the best."""
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_safe, 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_safe(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:]])
|
