Image-to-Video

UI build: 2026-05-09 (prompt field + Skip Wan if VRAM low) — if you do not see a Prompt box below, your Space is still running an old app.py; upload the latest file from your PC and Rebuild.

Upload an image, enter a prompt (optional but recommended on Wan), adjust duration/FPS, then click Generate Video.

Prompt: used when the Wan backend loads (large GPU). Stable Video Diffusion (smaller GPUs) moves the image but ignores text.

First run can take many minutes (downloading models). Watch Logs if the queue seems stuck.

GPU: use at least T4 or better in Space Settings.

Secrets: HF_TOKEN if downloads are gated; SKIP_WAN=1 if Wan hangs or OOM (forces SVD only).