---
license: apache-2.0
base_model:
  - Lightricks/LTX-2.3
tags:
  - video-generation
  - lora
  - ic-lora
  - ltx-video
  - camera-motion
  - lightricks
---

# LTX-2.3 22B IC-LoRA Any Trajectory Instruction (ATI)
<p align="center">
  <img src="assets/demo.gif"width="400" />
</p>

This is an experimental IC-Lora for porting [Any Trajectory Instruction (ATI)](https://github.com/bytedance/ATI) to LTX using minimal resources.

## Model Details

- **Base Model:** LTX-2.3-22b Video
- **Training Type:** IC LoRA
- **Control Type:** Video with splines in locations where motion tracking is given as a constraint
- **Bucket Size:** 768x768x81
- **Training Steps:** 3000
- **Learning Rate:** 1e-4
- **Dataset Size:** 25 video pairs
- **Hardware:** 1 H100
- **Training Time:** < 5 hours


## Using in ComfyUI
1. Download [ltx-2.3-22b-ic-lora-any-trajectory-instruction-3000.safetensors](./ltx-2.3-22b-ic-lora-any-trajectory-instruction-3000.safetensors) into `models/loras`.
2. Use the provided [IC-LoRA workflow](./ltx-2.3-22b-ic-lora-any-trajectory-instruction-workflow.json)
3. Draw motion paths on input image, ensure you use the following settings:

<img src="./assets/fl-path-animator-settings.png" alt="isolated" width="300"/>

## Limitations

- First experimental IC-LoRA checkpoint trained on only 25 video pairs
- Motion points occasionally appear on the results at different resolutions/frame count


## Acknowledgments

- LTX-Trainer: **https://github.com/Lightricks/LTX-2/tree/main/packages/ltx-trainer**
- ATI: **https://github.com/bytedance/ATI**
- FL Path Animator Node: **https://github.com/filliptm/ComfyUI_Fill-Nodes**

## Support
Consider supporting me if you want continuous updates and new IC-Loras!

[![buy-me-coffees](https://i.imgur.com/3MDbAtw.png)](https://www.buymeacoffee.com/yuvraj108cZ)
[![paypal-donation](https://i.imgur.com/w5jjubk.png)](https://paypal.me/yuvraj108c)