Train a MiniMax H3 LoRA on first/last/both keyframe signatures, teaching it to generate video with audio that starts on one image and lands on another.
No supported capabilities recorded.
/v1/models/fal/minimax/h3/flf2v/trainer