LTX 2.3 Reference-Guided Local Redraw & Video Inpainting Workflow

This workflow is designed for LTX 2.3 reference-guided local redraw and video inpainting. Its main purpose is to take a source video or control video, define the region that needs to be modified, and regenerate that part through LTX 2.3 while preserving the original timing, motion rhythm, audio, frame rate, and overall scene structure as much as possible.

Unlike a simple blur, crop, overlay, or single-frame repair method, this workflow is built as a video-level local reconstruction pipeline. The goal is not to destroy the whole shot and regenerate everything from scratch. Instead, the workflow keeps the source material as the structural guide, uses masks to define the editable area, and lets LTX 2.3 repaint only the required region while maintaining temporal continuity across frames. This makes it useful for authorized video cleanup, local object replacement, detail correction, character-area adjustment, controlled visual repair, and reference-guided video transformation.

The workflow uses LTX 2.3 as the main video generation backbone, with LTX video VAE, LTX audio VAE, LTXVAudioVAEEncode, LTXVConditioning, LTXVCropGuides, LTXVConcatAVLatent, LTXVSeparateAVLatent, SamplerCustomAdvanced, tiled VAE decoding, and final audio-video export logic. It also uses SetNode / GetNode routing for shared components such as FPS, total frames, duration, base model, video VAE, audio VAE, CLIP, control video, latent conditioning, masks, result frames, and audio. This modular structure makes the workflow easier to adapt to different videos and repeated local redraw tasks.

A key part of the workflow is the mask and guide system. The source video or control video is used to determine the frame size, frame count, and timing. SolidMask and related mask-routing logic define the region that should be affected. The workflow then crops and manages guide conditioning so that the model can focus on the intended local area instead of drifting across the full frame. This is important for video inpainting because uncontrolled redraw can easily change the face, clothing, background, camera angle, lighting, or motion outside the edited area.

The audio route is also preserved. The workflow includes audio loading / recording, audio VAE encoding, audio latent handling, and final audio routing. This means the final output can remain a complete video instead of becoming a silent visual-only repair. For creators working on AI video demonstrations, talking characters, short dramas, product clips, or social media edits, keeping the audio-video structure intact is a major practical advantage.

The workflow also uses multiple generation and refinement stages. The first stage builds the repaired latent video under the source guide and mask constraints. Later stages can continue refining the result, decode with tiled VAE logic, and prepare the final output. This makes the workflow more suitable for real production testing than a small single-pass experiment.

This workflow is ideal for creators who want to repair or modify a specific video region while keeping the rest of the shot stable. It can be used for LTX 2.3 local redraw tests, reference-guided video inpainting, AI video cleanup, controlled character-region edits, object-region repair, and Civitai / RunningHub workflow demonstrations. If you want to see how the control video, mask route, audio route, LTX 2.3 conditioning, cropped guides, and final local redraw output are connected, watch the full tutorial from the YouTube link above.

⚙️ Try the Workflow Online

👉 Workflow: https://www.runninghub.ai/post/2041676758285553666?inviteCode=rh-v1111

Open the link above to run the workflow directly online and view the generation results in real time.

If the results meet your expectations, you can also deploy it locally for further customization.

🎁 Fan Benefits: Register now to get 1000 points, plus 100 daily login points — enjoy 4090-level performance and 48 GB of powerful compute!

📺 Bilibili Updates (Mainland China & Asia-Pacific)

If you are in Mainland China or the Asia-Pacific region, you can watch the video below for workflow demos and a detailed creative breakdown.

📺 Bilibili Video: https://www.bilibili.com/video/BV1QrDzBsELY/

I will continue updating model resources on Quark Drive:

👉 https://pan.quark.cn/s/20c6f6f8d87b

These resources are mainly prepared for local users, making creation and learning more convenient.

⚙️ 在线体验工作流

👉 工作流： https://www.runninghub.ai/post/2041676758285553666?inviteCode=rh-v1111

打开上方链接即可直接运行该工作流，实时查看生成效果。

如果觉得效果理想，你也可以在本地进行自定义部署。

🎁 粉丝福利： 注册即送 1000 积分，每日登录 100 积分，畅玩 4090 体验 48 G 超级性能！

📺 Bilibili 更新（中国大陆及南亚太地区）

如果你在中国大陆或南亚太地区，可以通过下方视频查看该工作流的实测效果与构思讲解。

📺 B站视频： https://www.bilibili.com/video/BV1QrDzBsELY/

我会在 夸克网盘 持续更新模型资源：

👉 https://pan.quark.cn/s/20c6f6f8d87b

这些资源主要面向本地用户，方便进行创作与学习。