The fact that I can make videos like this with MiniMax H3 proves that MiniMax is the king of open-source video models. We're only scratching the surface of its true potential.
retained text
This video was generated at 0.6 megapixels, with 10 seconds per clip and 9 clips in total. It took around 2.5 hours to generate. After that, I upscaled it and applied 2x frame interpolation. I was inspired by Sinaboto's videos and wanted to create similar content for TikTok. After a lot of trial and error, I can finally make seamless videos like this with MH3! **Specs:** RTX 3060 12GB, 16GB RAM **Workflow:** [https://github.com/chanon/comfyui-obvpm-timeline](https://github.com/chanon/comfyui-obvpm-timeline) I can't wait to see what other people cooking, especially now that we've reached this level. I really hope they release an updated model, maybe MiniMax H3.5, with improvements to distant faces and objects, as well as action speed and motion consistency.
document fetched 2026-11-01 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xka3ps/the_fact_that_i_can_make_videos_like_this_with