# I managed to extend videos to any length in ComfyUI without losing consistency (Minimax-H3 + Visual Context Trick) — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/cf81f3de8d69bb5b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-09-11 17:39 ingested 2026-09-12 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1wdn30w/i_managed_to_extend_videos_to_any_length_in/)

## retained text



```
Hey everyone!

One of the biggest headaches with AI video has always been extending shots without the style degrading, characters morphing, or the cut being obvious.

I’ve been testing a method using **Minimax-H3** in ComfyUI to seamlessly cut and extend footage, and the results are honestly wild:

* **Pixel-perfect transitions:** The continuation aligns perfectly with the last frame of the original clip.
* **Context retention:** By feeding the model the visual context of the previous video, it actually remembers the specific assets (like the boat and character features) instead of hallucinating new ones.
* **Preserves aesthetic:** Keeps the lighting, colors, and overall camera style identical across cuts.

(Watch the preview clip to see the side-by-side transition!)

I’m currently packaging this into a **custom ComfyUI node** and recording a full walkthrough. Both the node and the full workflow will be **100% free** on my YouTube channel (**SatoDive**).  
[https://www.youtube.com/@SatoDive](https://www.youtube.com/@SatoDive)

Let me know what you think or if there are specific edge cases you’d like me to test before I release the tutorial!
```



document fetched 2026-09-12 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1wdn30w/i_managed_to_extend_videos_to_any_length_in
