--- tags: - lora - ic-lora - ltx-2.5 - video - video-editing - image-to-video - comfyui base_model: Lightricks/LTX-2.5 license: other library_name: comfyui pipeline_tag: video-to-video --- # LTX Ripple ### Edit one frame. Ripple the change through the whole video. LTX Ripple is a **First Frame All Frames (FFAF) IC-LoRA** for [LTX-2.5](https://huggingface.co/Lightricks/LTX-2.5). It propagates a visual edit made to the first frame across the rest of a video while preserving the source video's motion, timing, camera movement, composition, and unchanged content.

Created by WepeNerd

## What it does 1. Start with a video. 2. Edit or replace its first frame. 3. Run the provided workflow. 4. LTX Ripple carries the first-frame edit through the video. You can describe the edit in the prompt for additional guidance, but the LoRA can also work without an edit description. Useful applications include: - Special effects and environment changes - Style, lighting, and weather changes - Character attribute edits - Background replacement - Object removal or insertion - 3D model retexturing - Architectural visualization retexturing - Title or logo replacement - Sharpening and blurring ## Examples The example videos below are included in this repository. Open them in a new tab or download them to compare the propagated edits.
▶ ltx_001.mp4
Object replacement, effects, and lighting
▶ ltx_002.mp4
Lighting and background change
▶ ltx_003.mp4
Fire effect across cuts
▶ ltx_004.mp4
Liquid motion propagation
▶ ltx_005.mp4
Archviz / clay model texturing
▶ ltx_006.mp4
Sharpening and scene limitations
## Files | File | Purpose | |---|---| | `LTX25_Ripple_v11.safetensors` | LTX Ripple LoRA weights | | `Workflow_example/LTX_Ripple_FFAF_Edit.json` | Ready-to-import ComfyUI example workflow | | `Examples/` | Example output videos | ## Installation This is a **ComfyUI LoRA**. Download `LTX25_Ripple_v11.safetensors` and place it in: ```text ComfyUI/models/loras/LTX/LTX25_Ripple_v11.safetensors ``` Import `Workflow_example/LTX_Ripple_FFAF_Edit.json` into ComfyUI and update the model paths if your installation uses different locations. The workflow is built for LTX-2.5 and references the following model components: - [LTX-2.5 diffusion model](https://huggingface.co/Lightricks/LTX-2.5) - `ltx-2.5-video-vae-bf16.safetensors` - `ltx-2.5-audio-vae-bf16.safetensors` - `gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors` - `gemma4_e2b_it_int8_convrot.safetensors` — optional, only when prompt enhancement is enabled Required or referenced custom nodes: - [ComfyUI-LTXVideo](https://github.com/Lightricks/ComfyUI-LTXVideo) - [ComfyUI-KJNodes](https://github.com/kijai/ComfyUI-KJNodes) - [RES4LYF](https://github.com/ClownsharkBatwing/RES4LYF) - [ComfyUI-VideoHelperSuite](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite) - [cg-use-everywhere](https://github.com/chrisgoringe/cg-use-everywhere) ## Recommended settings The included workflow uses the following starting point: | Setting | Starting value | |---|---:| | LoRA model strength | `1.35` | | Frame rate | `25 fps` | | Video length | `97 frames` | | Guide frame index | `0` | | Guide strength | `1.0` | | Resolution | `1152 × 768` in the video-input branch | These are starting points, not hard requirements. Increase the LoRA strength when the edit is not being carried through strongly enough. Decrease it when the output changes content that should remain unchanged. ## Prompting The default prompt used by the workflow is: > Use the reference video for motion, timing, camera movement, composition, and unchanged scene content, while consistently propagating the visual edit established in the first frame throughout the video. For difficult edits, add a short description of the changed first frame. For example, describe the new lighting, object, material, style, or background that should remain consistent across the video. ## Limitations and tips - The first frame is the key control signal, so make the edit clear and coherent before running the video. - Stronger edits may need a higher LoRA strength and a more explicit prompt. - Very large changes, heavy occlusion, fast motion, or content absent from the source video may reduce consistency. - Results depend on the LTX-2.5 checkpoint, resolution, frame count, sampler settings, and available hardware. ## Credits LTX Ripple was created by [WepeNerd](https://huggingface.co/WepeNerd). It is designed for use with [LTX-2.5 by Lightricks](https://huggingface.co/Lightricks/LTX-2.5) and the [ComfyUI-LTXVideo](https://github.com/Lightricks/ComfyUI-LTXVideo) ecosystem. Please review and comply with the licenses and usage terms of LTX-2.5, its model components, ComfyUI, and the custom nodes listed above. ## License LTX Ripple is an LTX-2.x derivative and is distributed under the [LTX-2.x Community License Agreement](./LICENSE-2_x). The license governs use, modification, and redistribution of this LoRA when used with LTX-2.x. Please also review the [LTX Acceptable Use Policy](https://static.lightricks.com/legal/ltx-acceptable-use-policy.pdf).