LTX-2.3-22b-IC-LoRA-Outpaint Transforms Video Canvas Edges
Video creators can now expand existing footage using a new training module that fills designated blank areas with matching visual material. LTX-2.3-22b-IC-LoRA-Outpaint identifies pure black sections in an input clip and generates extensions that align with the original scene’s movement.
Developed by Oumoumad, this addition builds directly on the LTX-2.3-22b foundation model to handle canvas expansion tasks. The focused setup targets local processing needs by extending clips without altering existing frames.
Model Size: 1.31GB & VRAM GPU: requirements vary
Video canvas extension through targeted generation
- Reads pure black pixels as clear markers for areas that need new content.
- Supports extensions on any edge, including top, bottom, and side borders.
- Maintains visual alignment with the original footage during movement.
- Includes a ready-made setup for ComfyUI users.
- Functions independently through the LTXICLoRALoaderModelOnly node.
Local media teams can integrate this setup directly into existing pipelines to reframe shots without losing quality. Privacy-focused editors will appreciate the offline workflow, which keeps raw footage on personal drives while delivering consistent expansion results across different aspect ratios.
Practical steps for handling contrast and workflow integration
Users working with low-light footage should adjust exposure before starting the process. Deep shadows might blend into the required black padding, which can confuse the generation boundaries. A quick brightness adjustment resolves this issue.
"Before feeding to the model, apply gamma 2.0 (brightening) to your letterboxed input,"
noted the developer in a project page. After rendering, users simply reverse the adjustment to restore original lighting. The setup also runs through Wan2GP for alternative node environments.
Download the weights and workflows from the official repository.
