wording: no runtime post-processing

update demo (user edits)

drop (2B,best)/(780M,fast) annotations from model labels

models are public: drop HF_TOKEN logic

size labels: TEA-ASR-1 (2B) / TEA-ASR-1-mini (780M)

point demo at renamed repos TEA-ASR-1 / TEA-ASR-1-mini

pin gradio 5.49.1 to fix hub conflict

fix build: torch 2.8.0, python 3.12.12, module-level load

add example audio

TEA-ASR ZeroGPU gradio demo + example audio

initial commit