Voicebank : vraies voix françaises (CML-TTS) + pool anonyme + garde-fou Qwen3

Remplace la voicebank générée par Kokoro (timbre anglais sur français phonémisé -> accent que Qwen3 clonait) par 41 vraies voix FR issues de CML-TTS (livres audio studio) : 1 narrateur dédié, 18F/14M nommées, 4F/4M anonymes réservées. - scripts/import_voices.py : import multi-shards parquet, 1 clip/locuteur (le plus propre via levenshtein), genre estimé par F0 (YIN, anti-octave), filtre débit de parole (ref_text aligné sur l'audio). - VoiceEntry.anonymous + assign_voices : les figurants « anonyme (...) » tirent dans un pool réservé, jamais mélangé avec les voix nommées ; narrateur dédié (fr_narrator remplace fr_f_siwis). - dedup._anon_attrs : genre/âge déduits du nom anonyme (bon genre de voix). - tts/qwen3.py : garde-fou anti-dérive (rejette/réessaie les sorties en boucle ou coupées en estimant la durée plausible du chunk). Limite connue : Qwen3 ne sait pas synthétiser les fragments d'1-2 mots (incises, titres) -> trous ; à traiter (repli Kokoro ou fusion des incises). Inclut aussi du travail en cours antérieur (refacto backend LLM pluggable mlx/lmstudio, benchmark, ajustements frontend/API). Claude-Session: https://claude.ai/code/session_01XSVvcy1mfb4k1xDgib9vVU
2026-06-21 21:32:31 +02:00
parent 141df5f04e
commit ba1813c583
91 changed files with 2558 additions and 442 deletions
--- a/voicebank/clips/f_bella.wav
+++ b/voicebank/clips/f_bella.wav
--- a/voicebank/clips/f_emma.wav
+++ b/voicebank/clips/f_emma.wav
--- a/voicebank/clips/f_heart.wav
+++ b/voicebank/clips/f_heart.wav
--- a/voicebank/clips/f_nicole.wav
+++ b/voicebank/clips/f_nicole.wav
--- a/voicebank/clips/fr_anon_f_1.wav
+++ b/voicebank/clips/fr_anon_f_1.wav
--- a/voicebank/clips/fr_anon_f_2.wav
+++ b/voicebank/clips/fr_anon_f_2.wav
--- a/voicebank/clips/fr_anon_f_3.wav
+++ b/voicebank/clips/fr_anon_f_3.wav
--- a/voicebank/clips/fr_anon_f_4.wav
+++ b/voicebank/clips/fr_anon_f_4.wav
--- a/voicebank/clips/fr_anon_m_1.wav
+++ b/voicebank/clips/fr_anon_m_1.wav
--- a/voicebank/clips/fr_anon_m_2.wav
+++ b/voicebank/clips/fr_anon_m_2.wav
--- a/voicebank/clips/fr_anon_m_3.wav
+++ b/voicebank/clips/fr_anon_m_3.wav
--- a/voicebank/clips/fr_anon_m_4.wav
+++ b/voicebank/clips/fr_anon_m_4.wav
--- a/voicebank/clips/fr_f_1.wav
+++ b/voicebank/clips/fr_f_1.wav
--- a/voicebank/clips/fr_f_10.wav
+++ b/voicebank/clips/fr_f_10.wav
--- a/voicebank/clips/fr_f_11.wav
+++ b/voicebank/clips/fr_f_11.wav
--- a/voicebank/clips/fr_f_12.wav
+++ b/voicebank/clips/fr_f_12.wav
--- a/voicebank/clips/fr_f_13.wav
+++ b/voicebank/clips/fr_f_13.wav
--- a/voicebank/clips/fr_f_14.wav
+++ b/voicebank/clips/fr_f_14.wav
--- a/voicebank/clips/fr_f_15.wav
+++ b/voicebank/clips/fr_f_15.wav
--- a/voicebank/clips/fr_f_16.wav
+++ b/voicebank/clips/fr_f_16.wav
--- a/voicebank/clips/fr_f_17.wav
+++ b/voicebank/clips/fr_f_17.wav
--- a/voicebank/clips/fr_f_18.wav
+++ b/voicebank/clips/fr_f_18.wav
--- a/voicebank/clips/fr_f_2.wav
+++ b/voicebank/clips/fr_f_2.wav
--- a/voicebank/clips/fr_f_3.wav
+++ b/voicebank/clips/fr_f_3.wav
--- a/voicebank/clips/fr_f_4.wav
+++ b/voicebank/clips/fr_f_4.wav
--- a/voicebank/clips/fr_f_5.wav
+++ b/voicebank/clips/fr_f_5.wav
--- a/voicebank/clips/fr_f_6.wav
+++ b/voicebank/clips/fr_f_6.wav
--- a/voicebank/clips/fr_f_7.wav
+++ b/voicebank/clips/fr_f_7.wav
--- a/voicebank/clips/fr_f_8.wav
+++ b/voicebank/clips/fr_f_8.wav
--- a/voicebank/clips/fr_f_9.wav
+++ b/voicebank/clips/fr_f_9.wav
--- a/voicebank/clips/fr_f_siwis.wav
+++ b/voicebank/clips/fr_f_siwis.wav
--- a/voicebank/clips/fr_m_1.wav
+++ b/voicebank/clips/fr_m_1.wav
--- a/voicebank/clips/fr_m_10.wav
+++ b/voicebank/clips/fr_m_10.wav
--- a/voicebank/clips/fr_m_11.wav
+++ b/voicebank/clips/fr_m_11.wav
--- a/voicebank/clips/fr_m_12.wav
+++ b/voicebank/clips/fr_m_12.wav
--- a/voicebank/clips/fr_m_13.wav
+++ b/voicebank/clips/fr_m_13.wav
--- a/voicebank/clips/fr_m_14.wav
+++ b/voicebank/clips/fr_m_14.wav
--- a/voicebank/clips/fr_m_2.wav
+++ b/voicebank/clips/fr_m_2.wav
--- a/voicebank/clips/fr_m_3.wav
+++ b/voicebank/clips/fr_m_3.wav
--- a/voicebank/clips/fr_m_4.wav
+++ b/voicebank/clips/fr_m_4.wav
--- a/voicebank/clips/fr_m_5.wav
+++ b/voicebank/clips/fr_m_5.wav
--- a/voicebank/clips/fr_m_6.wav
+++ b/voicebank/clips/fr_m_6.wav
--- a/voicebank/clips/fr_m_7.wav
+++ b/voicebank/clips/fr_m_7.wav
--- a/voicebank/clips/fr_m_8.wav
+++ b/voicebank/clips/fr_m_8.wav
--- a/voicebank/clips/fr_m_9.wav
+++ b/voicebank/clips/fr_m_9.wav
--- a/voicebank/clips/fr_narrator.wav
+++ b/voicebank/clips/fr_narrator.wav
--- a/voicebank/clips/m_eric.wav
+++ b/voicebank/clips/m_eric.wav
--- a/voicebank/clips/m_fenrir.wav
+++ b/voicebank/clips/m_fenrir.wav
--- a/voicebank/clips/m_george.wav
+++ b/voicebank/clips/m_george.wav
--- a/voicebank/clips/m_lewis.wav
+++ b/voicebank/clips/m_lewis.wav
--- a/voicebank/clips/m_michael.wav
+++ b/voicebank/clips/m_michael.wav
--- a/voicebank/clips/m_santa.wav
+++ b/voicebank/clips/m_santa.wav