FIX: wrap inference in @spaces.gpu (was CPU fallback -> hollow noise); add sampling+normalize

fix README short_description <=60 chars

drop pycache, add gitignore

batch Parler-TTS: multi-set JSON -> zip of audios + manifest (zero-gpu)

add zero-gpu hardware spec + batch UI

rebuild: Parler-TTS batch engine

Pin torch and torchaudio to 2.1.2 to avoid torchcodec error

Remove unidic to fix MeCab error

Update requirements to fix BeamSearchScorer error

Fix transformers version mismatch for XTTS v2

Fix 503 crash: Add TOS agreement and auto-download sample voice

Add app.py and requirements

initial commit