LTX-2.3 · Motion Transfer (Pose)

ABOUT THE WORKFLOW

Transfer motion to any subject
Upload a reference video and the workflow reads its body movement. Give it a subject to animate, either a photo you upload or one you describe in the prompt, and the model performs the same motion with that subject. The audio from the reference clip carries over into the result.

Model

• LTX-2.3 (22B) by Lightricks. An open weights video model that generates picture and sound together. It runs here with a Union Control adapter for pose and a distilled speed adapter that keeps the step count low. Released under the LTX-2 Community License.

HOW IT WORKS

Step 1. Upload your reference video
The clip whose movement you want. Its body motion drives the result.
Works great with: dance clips · full-body movement · single-person action

Step 2. Add a subject photo (optional)
A portrait of the person or character to animate. Skip it to build the subject from your prompt instead.

Step 3. Write your prompt
Describe the subject and the scene. With a photo, it sets the setting and style. Without one, it defines who appears. A negative prompt comes pre-filled to keep out game and cartoon looks.

Step 4. Hit run and download
The model animates the subject to match the motion, with the reference audio carried over. This is a large model, so give it a little longer to finish.
Ready for: editing timelines · social · previs

First time? Leave every setting as-is. The defaults are the right starting point for almost everyone.

RECOMMENDED SETTINGS

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

• Standard use (most people) — 960x544, 121 frames, 8 steps, guidance 1, distilled strength 0.5. The right starting point for almost everyone.

• No photo, prompt only — skip the subject image and turn on bypass_i2v. The subject then comes from your prompt while the reference video still drives the motion.

• Want a longer clip — raise the frame count above 121. At 24 frames per second, 121 frames is about five seconds, so add frames for more time.

• Want more detail in the result — lower the distilled strength below 0.5. It leans less on the speed adapter, which adds detail and takes longer.

• Want a different take — change the seed. Keep the same seed to reproduce a run you liked.

• The motion looks off — use a reference video with clear, full-body movement and a single person. Cluttered or multi-person clips confuse the pose read.

Prompt: With a photo, describe the scene, wardrobe, and mood so the look holds together while the movement comes from your reference video. Without a photo, describe the subject too, like "a woman in a red dress dancing in a neon-lit studio" rather than "dancing." The negative prompt already lists game and cartoon looks to avoid.

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

🎭 Character Animation
Animate an uploaded portrait with a real performance pulled from a reference clip, audio and all.

✍️ Prompt to Motion
Skip the photo and generate a subject straight from text, then drive it with reference footage.

🕺 Dance & Performance
Map a dance or movement clip onto your subject and keep the reference audio in the cut.

🎬 Previs & Storyboarding
Test how a shot moves by driving a subject with reference footage before you shoot it.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• A clear, front-facing subject photo when you upload one

• Reference clips with a single person

• Full-body, well-lit movement in the reference

• Prompts that name the subject, setting, and mood

⚠️ May produce softer results

• Crowded or multi-person reference clips

• Fast, blurry, or partly cropped motion

• Side profiles or obscured faces in the subject photo

• Low-resolution or clipped reference video

FAQ

What is LTX-2.3?
LTX-2.3 is an open weights audio-video model from Lightricks. It generates video and synchronised sound together in a single pass. This workflow uses the 22B checkpoint, released under the LTX-2 Community License, with adapters for pose control and faster generation.

How does motion transfer work in this workflow?
The workflow reads the body movement from your reference video using pose estimation, then applies that pose to your subject through a Union Control adapter. The subject can come from a photo you upload or from your text prompt. LTX-2.3 then generates a new video of that subject following the same motion.

Can I use this without a subject photo?
Yes. Skip the image and turn on bypass_i2v, then describe the subject in your prompt. The reference video still supplies the motion, so the model builds the subject from your words and animates it with the movement from the clip.

Does the output include audio?
Yes. The audio from your reference clip carries over into the result, and LTX-2.3 generates sound alongside video by design. That means the final clip comes back with a matching audio track rather than a silent render you have to score later.

Can I use LTX-2.3 commercially?
LTX-2.3 is released under the LTX-2 Community License. It is free for commercial use if your company earns under 10 million dollars in annual revenue, and Lightricks offers a separate commercial license above that. Make sure you also hold the rights to any photo or video you upload.

What is the Union Control adapter, and how is pose mode different from canny or depth?
The Union Control adapter lets one model follow pose, canny edges, or depth as its guide. This workflow runs in pose mode, so it follows the body movement in your reference video rather than its outlines or its distance from the camera. That keeps the focus on how the subject moves.

How to run LTX-2.3 online?
You can run LTX-2.3 online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, upload your reference video, add a photo or a prompt, 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.

An animator runs a motion transfer and likes the take. 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 reference video, add a photo or write a prompt, then run it. The settings are already set.

→ Launch Workflow, Free

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