Lora Based Motion Concept Learning From Video Reference Sets

via “camera control and motion specification through ic-lora”

LTX-Video Support for ComfyUI

Unique: Implements IC-LoRA conditioning system that enables camera and motion control without full model retraining. Integrates with LTXVQ8LoraModelLoader to support quantized IC-LoRA weights, enabling efficient motion-controlled generation on memory-constrained systems.

vs others: More precise camera control than text-only prompts; enables reproducible camera movements across multiple generations, unlike prompt-based approaches which produce variable results.

via “lora-based motion concept learning from video reference sets”

[ECCV 2024 Oral] MotionDirector: Motion Customization of Text-to-Video Diffusion Models.

Unique: Implements dual-path LoRA decomposition (spatial vs temporal) enabling independent training and composition of appearance and motion, rather than monolithic fine-tuning. Uses selective LoRA injection only into temporal attention/cross-attention layers, preserving spatial reasoning from base model while learning motion dynamics.

vs others: More parameter-efficient than full fine-tuning (0.5-2% of model parameters) and faster than DreamBooth-style approaches, while maintaining better motion fidelity than simple prompt engineering or classifier-free guidance alone.