LTX-2.3 22B IC-LoRA Deinterlace

This is a Deinterlace IC-LoRA trained on top of LTX-2.3-22b, designed to remove temporal artifacts found in poorly converted or transcoded footage: interlace combing (true interlaced material left progressive, or shown with the wrong field order), bad linear/bilinear deinterlacing, and frame-blending ghosting from bad FPS conversion (frame averaging, mistimed interpolation). The model takes the corrupted clip as the conditioning input and tries to reconstruct a clean progressive version.

It is based on the LTX-2.3 foundation model.

⚠️ Status: still not yet well tested. Treat this checkpoint as a preview — quality on real-world degraded footage hasn't been thoroughly validated yet. Feedback welcome.

Model Files

ltx-2.3-22b-ic-lora-deinterlace.safetensors

Model Details

• Base Model: LTX-2.3-22b
• Training Type: IC LoRA
• Purpose: Remove interlace combing, wrong-field-order judder, bad-deinterlace softening, and frame-blending ghosting from input video
• Training Steps: 5000

🔌 Using in ComfyUI

1. Copy the LoRA weights into models/loras.
2. Use the IC-LoRA workflow from the LTX-2 ComfyUI repository.
3. Load the LoRA using the LTXICLoRALoaderModelOnly node.

License

See the LTX-2-community-license for full terms.