Run LTX-Video 0.9.7-distilled locally on ZeroGPU instead of proxying to H3

Drop @spaces.GPU from the remote-Space fn; retry + readable errors

Drop @spaces.GPU from the remote-Space fn; retry + readable errors

Update requirements.txt

Update app.py

Update app.py

Use MiniMax-H3 via gradio_client (33B model too big for local ZeroGPU)

Use MiniMax-H3 via gradio_client (33B model too big for local ZeroGPU)

Use MiniMax-H3 via gradio_client (33B model too big for local ZeroGPU)

Remove hf_oauth so the demo runs anonymously (no sign-in gate)

Lower @spaces.GPU duration to 60s (180->270s exceeded visitor ZeroGPU max)

Video fn output: return {path, url:data-uri} so the canvas player can load it (bare path = local, unloadable)

Enable hf_oauth (per-user ZeroGPU quota)

Add imageio for browser-compatible H.264 mp4 export

Add opencv-python-headless for export_to_video

Use LTXImageToVideoPipeline (image-to-video)

Load pipeline at startup (correct ZeroGPU pattern); 49 frames

ZeroGPU animator: fn node + @spaces.GPU + LTX-Video

ZeroGPU animator: fn node + @spaces.GPU + LTX-Video

ZeroGPU animator: fn node + @spaces.GPU + LTX-Video

ZeroGPU animator: fn node + @spaces.GPU + LTX-Video

initial commit