Uploaded April 2026 | Updated November 2026, 2 days ago
I built this LTX 2.3 in-painting workflow for ComfyUI so you can easily edit any video. Most editing methods ruin your original background, but our three-pass setup prevents that completely. You run phase one to build the basic structure, and then phases two and three clean up the fine details and complex errors. The original camera angle and lighting remain exactly the same while your specific object or person changes.
I included two different versions for you. Use the first one when you have a specific reference image. Use the second version when you only want to change the video using a text prompt.
You must download the official masked inpainting LoRA and save it directly into your ComfyUI LoRA folder. You can also stack the new distilled LoRA directly on top of your base model if you want an upgraded visual style and better audio.
Writing precise prompts is difficult, so I integrated Qwen-VL to do the heavy lifting for us. It reads your source video and uploaded image, then automatically writes your first draft. You just read the generated text and manually fix the motion instructions if the AI makes a mistake.
If your computer has low VRAM, simply switch to the GGUF option. I recommend dropping your scale setting from 1 down to 0.5 to protect the original structure of your shot. Also, if your source video lacks sound, the generation will crash. We fix this by connecting the line directly to the Empty Latent Audio node.
I set the default masking block size to 32 to give us the perfect balance. A high block size makes a stable mask but loses edge detail, while a low block size tracks tiny details but becomes highly sensitive to errors.
Watch my full step-by-step guide to building this node setup: youtube.com/watch?v=lQcter3uVQA&t=165s
LTX 2.3 LoRA: 5 Ways to Edit ANY Video With Text: youtube.com/watch?v=2piBFWVIUbs
LTX 2.3 ComfyUI: 3-Stage 4K Workflow: youtube.com/watch?v=7qLPvI4Wkko
LTX 2.3 ComfyUI Workflow: The ID-LoRA Perfect Lip Sync Fix: youtube.com/watch?v=-NPcoFF6QkU
LTX 2.3 ComfyUI Workflow: The ID-LoRA Perfect Lip Sync Fix: youtube.com/watch?v=-NPcoFF6QkU
Workflow Link: aistudynow.com/how-i-built-the-perfect-ltx-2-3-in-painting-comfyui-workflow
I built this LTX 2.3 in-painting workflow for ComfyUI so you can easily edit any video. Most editing methods ruin your original background, but our three-pass setup prevents that completely. You run phase one to build the basic structure, and then phases two and three clean up the fine details and complex errors. The original camera angle and lighting remain exactly the same while your specific object or person changes.
I included two different versions for you. Use the first one when you have a specific reference image. Use the second version when you only want to change the video using a text prompt.
You must download the official masked inpainting LoRA and save it directly into your ComfyUI LoRA folder. You can also stack the new distilled LoRA directly on top of your base model if you want an upgraded visual style and better audio.
Writing precise prompts is difficult, so I integrated Qwen-VL to do the heavy lifting for us. It reads your source video and uploaded image, then automatically writes your first draft. You just read the generated text and manually fix the motion instructions if the AI makes a mistake.
If your computer has low VRAM, simply switch to the GGUF option. I recommend dropping your scale setting from 1 down to 0.5 to protect the original structure of your shot. Also, if your source video lacks sound, the generation will crash. We fix this by connecting the line directly to the Empty Latent Audio node.
I set the default masking block size to 32 to give us the perfect balance. A high block size makes a stable mask but loses edge detail, while a low block size tracks tiny details but becomes highly sensitive to errors.
Watch my full step-by-step guide to building this node setup: youtube.com/watch?v=lQcter3uVQA&t=165s
LTX 2.3 LoRA: 5 Ways to Edit ANY Video With Text: youtube.com/watch?v=2piBFWVIUbs
LTX 2.3 ComfyUI: 3-Stage 4K Workflow: youtube.com/watch?v=7qLPvI4Wkko
LTX 2.3 ComfyUI Workflow: The ID-LoRA Perfect Lip Sync Fix: youtube.com/watch?v=-NPcoFF6QkU
LTX 2.3 ComfyUI Workflow: The ID-LoRA Perfect Lip Sync Fix: youtube.com/watch?v=-NPcoFF6QkU
Workflow Link: aistudynow.com/how-i-built-the-perfect-ltx-2-3-in-painting-comfyui-workflow










