Live Conversation + Backyard AI submission: tags, demo videos, social link

Avatar: mouse-look, particles, bloom, cyberpunk HUD

Tutor: reliable pre-generated audio (gTTS static files, no GPU)

Tutor: rotating phrases, direct scoring, best-effort audio (conserve GPU)

Tutor: log why synth returns no audio

Tutor: speak phrase + feedback via Seamless (any language), autoplay on

Fix Seamless: processor audio= kwarg (was audios=), duration=60

Fix build: unpin platform packages, pin gradio sdk_version 5.38.0

Deploy Reachy Bridge interpreter (SeamlessM4T on ZeroGPU)

initial commit