Add pinned local Qwen GPU voice generation and first puzzle drafts

tools/audio/prepare_voice.py

@@ -0,0 +1,73 @@
+"""Resample validated source WAVs into draft Unity clips without claiming listening approval."""
+import argparse
+from pathlib import Path
+import math
+import sys
+
+import numpy as np
+import soundfile as sf
+from scipy.signal import resample_poly
+
+from generate_voice import load_json, validate_manifest, waveform_metadata, write_json
+
+
+def prepare(manifest, catalog, profile, root, target, ids=None):
+    errors = validate_manifest(manifest, catalog, profile, root, ids)
+    if errors:
+        raise ValueError('; '.join(errors))
+    selected = set(ids or [c['id'] for c in catalog['clips'] if c['required']])
+    for entry in manifest['clips']:
+        if entry['id'] in selected and entry['id'].startswith('count_') and entry['duration'] > 2:
+            raise ValueError(f"{entry['id']}: count utterance exceeds 2 seconds; review before runtime export")
+    target.mkdir(parents=True, exist_ok=True)
+    edited_dir = root / 'edited' / manifest['runId']
+    edited_dir.mkdir(parents=True, exist_ok=True)
+    selections = []
+    for entry in manifest['clips']:
+        if entry['id'] not in selected:
+            continue
+        wave, rate = sf.read(root / entry['file'], dtype='float32')
+        divisor = math.gcd(rate, 44100)
+        edited = resample_poly(wave, 44100 // divisor, rate // divisor)
+        if np.max(np.abs(edited)) >= 1:
+            raise ValueError(f"{entry['id']}: editing would clip; review level before export")
+        edited_path = edited_dir / (entry['id'] + '.wav')
+        sf.write(edited_path, edited, 44100, subtype='PCM_16')
+        runtime_path = target / edited_path.name
+        if runtime_path.exists() and runtime_path.read_bytes() != edited_path.read_bytes():
+            raise ValueError(f"{entry['id']}: existing runtime clip preserved; review new edited candidate before replacing it")
+        runtime_path.write_bytes(edited_path.read_bytes())
+        selections.append({'id': entry['id'], 'sourceRunId': entry['originalRunId'],
+                           'sourceFile': entry['file'], 'sourceSha256': entry['sha256'],
+                           'editedFile': edited_path.relative_to(root).as_posix(),
+                           'runtimeFile': 'Assets/MagnetismHeroes/UnityRuntime/Presentation/Audio/Voice/' + runtime_path.name,
+                           'editedMetadata': waveform_metadata(edited_path),
+                           'review': 'not_listened', 'status': 'draft',
+                           'processing': 'polyphase resample to mono PCM16 44100 Hz; no speech trimming'})
+    result = {'schemaVersion': 1, 'runId': manifest['runId'], 'profileHash': manifest['profileHash'],
+              'catalogHash': manifest['catalogHash'], 'selections': selections}
+    write_json(edited_dir / 'selection.json', result)
+    return result
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--catalog', type=Path, required=True)
+    parser.add_argument('--profile', type=Path, required=True)
+    parser.add_argument('--manifest', type=Path, required=True)
+    parser.add_argument('--ids')
+    parser.add_argument('--target', type=Path, default=Path('Assets/MagnetismHeroes/UnityRuntime/Presentation/Audio/Voice'))
+    args = parser.parse_args()
+    try:
+        selection = prepare(load_json(args.manifest), load_json(args.catalog), load_json(args.profile),
+                            args.catalog.parent, args.target, args.ids.split(',') if args.ids else None)
+        for clip in selection['selections']:
+            print(f"{clip['id']}: draft {clip['editedMetadata']['duration']:.3f}s {clip['editedMetadata']['sha256']}")
+        return 0
+    except Exception as error:
+        print(str(error), file=sys.stderr)
+        return 1
+
+
+if __name__ == '__main__':
+    sys.exit(main())