LTX 2.3 IC-LoRA Video Colorization Workflow

This workflow is designed for LTX 2.3 IC-LoRA video colorization with manual prompt control. Its main purpose is to take a video or frame-guided video input, preserve the original motion and structure, and use LTX 2.3 with an IC-LoRA Colorizer route to rebuild the clip with stronger color, more coherent lighting, and a more cinematic final look.

Unlike a simple video filter or LUT-based color correction workflow, this setup does not only push saturation or contrast on top of the original frames. It uses LTX 2.3 generation, video latent conditioning, image-to-video guidance, IC-LoRA guide injection, prompt-based color direction, tiled decoding, and audio-video latent routing to regenerate the video while keeping the underlying motion consistent. This makes it more useful for AI video restoration, grayscale-to-color tests, stylized recoloring, cinematic color enhancement, and controlled visual remastering.

The workflow uses ltx-2.3-22b-dev as the main video model route, combined with ltx-2.3-22b-distilled-lora-384 and LTX-2.3-22b-IC-LoRA-Colorizer-0.9. It also includes LTX video VAE, LTX audio VAE, Gemma / LTX text encoding, LTXVPreprocess, EmptyLTXVLatentVideo, LTXVImgToVideoConditionOnly, LTXAddVideoICLoRAGuide, LTXVConcatAVLatent, LTXVSeparateAVLatent, LTXVCropGuides, SamplerCustomAdvanced, VAEDecodeTiled, and final audio-video export logic.

The most important part of this workflow is the IC-LoRA color guide. LTXAddVideoICLoRAGuide injects the guide image or video reference into the latent process, allowing the model to follow the original structure while applying the intended color and visual direction. This is useful when you want the final video to retain the same movement, framing, and scene layout, but gain richer color, stronger material separation, more realistic lighting, or a more stylized cinematic palette.

The workflow is also built for manual prompt control. Instead of fully relying on automatic enhancement, the creator can describe the target color style directly: realistic skin tones, warm sunlight, neon cyberpunk lighting, dark fantasy grading, natural outdoor colors, filmic contrast, or any other visual direction. This gives the workflow more creative flexibility than a fixed colorization model.

The graph also includes structure-control tools such as DepthCrafter, Canny edge preprocessing, and DW pose preprocessing routes. These help preserve spatial depth, edge structure, and body motion when needed. For video colorization, this matters because color changes should not break the original motion, deform the subject, or destabilize the scene.

The workflow keeps an audio-video structure through LTXVEmptyLatentAudio, LTXVConcatAVLatent, LTXVSeparateAVLatent, and LTXVAudioVAEDecode. This means the final output can remain a complete video instead of becoming a silent frame sequence. Tiled VAE decoding is used to make the final video decoding more stable and practical for larger frames.

This workflow is ideal for AI video colorization, black-and-white video remastering, stylized recoloring, cinematic grading tests, old-video enhancement, character video cleanup, LTX 2.3 IC-LoRA experiments, RunningHub demos, and Civitai workflow previews. If you want to see how LTX 2.3, IC-LoRA Colorizer, manual prompt control, depth / edge / pose guidance, and final video export work together, watch the full tutorial from the YouTube link above.

⚙️ Try the Workflow Online

👉 Workflow: https://www.runninghub.ai/post/2035643489333022722?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/BV11sAGzvE1M/

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/2035643489333022722?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/BV11sAGzvE1M/

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。