LTX 2.3 IC LoRA Union Control · V2V For Anime

WORKFLOW PAGE COPY

LTX 2.3 IC LoRA Union Control · Video to Video

Title: Restyle Any Video with LTX 2.3 IC LoRA Union Control

Subtitle: Upload a video and a style reference image. LTX 2.3 with IC LoRA Union Control restyles the full video to match the reference while preserving the original motion, body pose, and scene structure using blended depth, pose, and edge control.

Tags: ltx 2.3, ic lora, union control, video to video, style transfer, restyle, structural control, open source

ABOUT THE WORKFLOW

Restyle a Video with Structural Control
Upload a reference video and a style reference image. The workflow extracts three structural maps from every video frame (depth, body pose, and edge detection), blends them into a single control signal, and feeds them alongside the style image into LTX 2.3 with IC LoRA Union Control. The LoRA combines image conditioning (visual style from the reference) with structural control (motion and layout from the video) in a single adapter. The original audio carries through. The output is a 1080p MP4 at 24fps.

Model

• LTX Video 2.3 22B Dev by Lightricks with IC LoRA Union Control. A single LoRA that unifies image conditioning and structural control (depth, pose, edge) so the model follows both the style reference and the video's motion simultaneously.

• DepthAnythingV2, DWPreprocessor (OpenPose), and Canny Edge. Three preprocessors that extract structural maps from the source video, blended at 0.75 weight for combined control.

• LTX 2.3 Distilled LoRA (strength 0.5). Speed adapter for 8-step generation.

HOW IT WORKS

Step 1. Upload your reference video
The video you want to restyle. The workflow extracts frames and audio, then generates structural control maps from every frame.
Works great with: live-action footage · AI-generated clips · anime · stock video

Step 2. Upload your style reference image
A single image that defines the target visual style. The IC LoRA reads the colors, textures, art direction, and rendering style from this image and applies them across every frame of the video.
Works great with: anime screenshots · illustration styles · painterly references · cinematic stills

Step 3. Write your prompt
Describe the target style for the restyle. "Convert the input video into the anime style shown in the input image, all characters should be anime kind" tells the model what aesthetic to apply. Match the prompt to the style reference.

Step 4. Hit run and download
The workflow generates at 960x544 with blended structural control, decodes with tiled VAE, resizes to 1920x1080, and combines with the original audio.
Ready for: Premiere · DaVinci Resolve · After Effects · TikTok · YouTube

First time? Upload a video and a style reference image. Edit the prompt to describe the style. Leave everything else as-is.

RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

• Standard video restyle — Upload video + style image, 960x544 internal, 121 frames, 8 steps, CFG 1. Write a style prompt and run.

• Stronger style transfer — Raise the IC LoRA guide strength from 0.75 toward 1.0. The output follows the style reference more aggressively at the cost of some structural fidelity.

• Closer to the original video structure — Lower the IC LoRA guide strength toward 0.5. The output stays structurally closer to the source video with lighter style application.

• Different control emphasis — The three control maps (depth, pose, canny) are blended at 0.75 weight. Adjust the ImageBlend ratio to emphasize depth over edges or vice versa.

• Longer clip — Increase the EmptyLTXVLatentVideo frame count beyond 121. More frames at 24fps extends the duration. VRAM usage scales with length.

• Flickering between frames — Lower the distilled LoRA strength from 0.5 toward 0.3. Or try a different seed. The structural control blend helps temporal consistency, but fast-moving source video may still flicker.

• Reproduce a result — The seed defaults to fixed (564654654). Keep it locked for identical output.

Prompt: Describe the target style, not the scene content. "Convert to anime style with cel shading, clean lineart, vibrant colors" works when the style reference is anime. "A person walking down a hallway" describes the video content, which the model already reads from the structural maps.

LEARN

📹 Videos

• Intro to Floyo

• ComfyUI 101 Free Course ft. Sebastian Kamph

• Floyo 101 for Team Collaboration

✨ Quick links

• Help docs

• LoRA Training guide

• Discover more workflows

USE CASES

🎨 Full Video Style Transfer
Restyle live-action footage into anime, illustration, painterly, or any target aesthetic defined by a single reference image, while preserving the original motion and audio.

🎬 Look Development and Previsualization
Test visual styles on existing footage before committing to full production. Swap the style reference image to compare different aesthetics on the same source video.

📱 Social Content Repurposing
Transform stock footage, screen recordings, or existing clips into stylized content for TikTok, Reels, and Shorts without reshooting.

🎮 Game and Animation Reference
Convert live-action reference clips into the target art style for use as animation guides, storyboard previews, or style bibles.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Style reference images with a clear, distinct aesthetic (anime screenshot, painterly illustration)

• Videos with visible subjects and consistent lighting

• Moderate motion where depth, pose, and edges are clearly readable

• Prompts that describe the style, not the scene content

⚠️ May produce softer results

• Style references with a generic or ambiguous aesthetic

• Videos with rapid scene cuts or extreme camera shake

• Very dark footage where depth and edge detection are unreliable

• Prompts that contradict the style reference image

FAQ

What is IC LoRA Union Control?
IC LoRA Union Control is a single LoRA adapter for LTX 2.3 that combines two capabilities: image conditioning (reading visual style from a reference image) and structural control (following depth, pose, and edge maps from a source video). Standard workflows require separate ControlNets or adapters for each. Union Control merges them into one LoRA so the model follows both the style target and the video structure simultaneously.

What structural control maps does this workflow extract?
Three types, blended into a single control signal: DepthAnythingV2 for spatial layout, DWPreprocessor (OpenPose) for body, hand, and face keypoints, and Canny Edge for line structure. The blend ratio (default 0.75) controls the relative weight between them.

Does the restyle preserve the original audio?
Yes. The workflow extracts audio from the source video and embeds it in the restyled output. The audio passes through unchanged.

How is this different from the Wan 2.1 Fun Control V2V restyle workflow?
This workflow uses LTX 2.3 with a single IC LoRA that handles both style and control in one adapter. The Wan 2.1 Fun Control workflow uses a separate FLUX model to generate the style reference and a separate control selection (depth OR scribble OR pose). LTX Union Control blends all three control types together and reads the style from an uploaded image rather than generating one.

Can I use my own style reference image?
Yes. Upload any image that represents the target visual style: an anime screenshot, an oil painting, a pixel art sprite, or a photographic look. The IC LoRA reads the aesthetic from this single image and applies it across the full video.

Is LTX Video 2.3 licensed for commercial use?
LTX Video 2.3 is an open-source model by Lightricks. The IC LoRA Union Control and other LoRAs have their own license terms. Check each component's license on its model page for commercial use in your specific project.

How to run LTX 2.3 IC LoRA video restyle online?
You can run LTX 2.3 IC LoRA video restyle online through Floyo. No installation, no setup, no local GPU needed. Open the workflow in your browser, upload your inputs, and hit run. Free to try.

WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

A designer runs an edit and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.

Ready to try it?
Upload a video and a style reference image, describe the target style, and hit run.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.