Workflow：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2064244173393457153/?inviteCode=rh-v1497">https://www.runninghub.ai/post/2064244173393457153/?inviteCode=rh-v1497</a>Sign up through my link to receive 1,000 credits, then claim 100 credits with each daily login.YouTube tutorial：<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/0tJ8inVkiBI" start="0"></iframe></div>This workflow is based on LTX 2.3 and integrates the Licon MSR custom node developed by Bilibili creator Light Cone Lab, designed to tackle the notoriously difficult problem of multi-subject consistency in AI video generation. Unlike the traditional LTX tendency toward artistic or slow-motion styles, after deep tuning it can stably output fast-paced martial arts combat and other high-speed action scenes without any post-production speed-up. The core idea is to pack 2 to 5 static reference images into a "pseudo-video" that acts as the model's visual memory, allowing the AI to accurately remember and reproduce multiple independent visual elements. The 50fps high-frame-rate setting further smoothens the transition of fast-moving object details, while an integrated large language model intelligently expands action prompts, effectively avoiding monotonous motion in high-speed scenes.Core Features:- Integrated Licon MSR node, packing multiple static reference images into a guiding video to achieve multi-subject consistency- Supports 50fps high-frame-rate output, eliminating ghosting and choppiness for fast-moving objects- Can blend 2–5 different reference images, enabling multiple specific characters to appear and interact spatially within a single video- Integrates a large language model (LLM) to expand action descriptions in prompts, maintaining motion complexity and plausibility in high-speed scenes- Requires at least two input images (first frame and background image); the background image is critical for scene stability

LTX 2.3 High-Motion Multi-Reference Workflow

ltx23HighMotionMulti_highMotionLTX23.zip

High-Motion LTX 2.3

RunningHub workflow：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2065361012815122434/?inviteCode=rh-v1497">https://www.runninghub.ai/post/2065361012815122434/?inviteCode=rh-v1497</a>Sign up through my link to receive 1,000 credits, then claim 100 credits with each daily login.YouTube tutorial：<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/amx-_5zoksM" start="0"></iframe></div>🌟 OverviewSay goodbye to massive workflows packed with masks, pose detection, alignment nodes, and separate tools.Powered by Wan 2.2, this workflow brings ByteDance’s unified Bernini framework into one all-in-one AI video editor. Edit videos with text, replace characters, transfer motion, restyle scenes with reference images, insert new content, and generate coherent videos from multiple images—all inside one workflow.🚀 Core Features• V2V Video EditingChange weather, remove objects, replace backgrounds, or modify actions with a single prompt.• Reference-Guided Video EditingUse reference images to restyle scenes, replace characters, and transfer motion without masks or pose detection.• Content InsertionPlace images or video clips onto billboards, screens, and other tracked surfaces.• Multi-Image Video GenerationCombine up to five reference images into a coherent HD video with smooth motion and camera movement.• Automatic Prompt EnhancementTurn simple instructions into detailed prompts using the Bernini Prompt Enhancer and an online or local LLM.⚙️ Workflow NotesBuilt on Wan 2.2 with WanAnimatePlus ModelLoader and acceleration LoRA. Lower-VRAM users can increase BlackSwap to around 30–40. Reference images and videos can be enabled or disabled with Ctrl+B.

Files with SHA256: 1a334074ddbb33a0bef1eb6db30ff8190cc2ac9aac60c4c72b9e89f2cdc4c33b