Image To Video Animation

via “image-to-video animation with conditioning frames”

Official repository for LTX-Video

Unique: Implements multi-position frame conditioning through latent-space injection at arbitrary temporal indices, allowing precise control over which frames match input images while diffusion generates surrounding frames, vs. simpler approaches that only condition on first/last frames

vs others: Supports arbitrary keyframe placement and multiple conditioning frames simultaneously, providing finer temporal control than Runway's image-to-video which typically conditions only on frame 0