Remove temporary ZeroGPU probe

Add isolated ZeroGPU CUDA probe

Remove obsolete Gradio 6 launch option

Update gallery for Gradio 6

Use 10 second ZeroGPU large allocation

Test ZeroGPU large allocation

Convert Space to Higgs Audio v3 only

Update app.py

Auto-transcribe selected voice with CPU-only Whisper to pre-fill the Higgs reference transcript

Add Higgs Audio v3 (4B) as a selectable voice backend alongside DramaBox

Fix: use <audio> HTML in card instead of gr.Audio (filepath rejects URLs)

Rewrite gallery: native Gradio card grid, real Use buttons, working filters/pagination

Bundle voices.json locally (11430 voices); load from file not URL

Fix: gr.Markdown does not accept scale kwarg

Voice gallery UI — enumerate el_voices.json with filter/preview/select flow

Initial DramaBox TTS space