Update README.md

Update requirements.txt

Live mode: single continuous ZeroGPU session generator (SDXL-Turbo pattern); buffer all mic audio, drop nothing

Fix Live mode: keep session-id bridge in DOM (CSS-hide) so prep_live_reference result reaches the JS and enables Start

Add progress indicators; disable Live Start until reference ready

Live mode: use non-streaming gr.Audio for target-speaker reference

Add additive live mode: custom HTML mic-streaming voice conversion with 60s ZeroGPU session + friendly continue prompt

Update README.md

Calibrate ZeroGPU duration from measured runs; int16 output; Gradio 6 theme/css on launch()

MeanVC2 zero-shot voice conversion demo

initial commit