LTX 2.5 is releasing today, and from one of the example videos I’ve seen, face distortion still doesn’t seem fully fixed.
That’s already a weak spot for LTX. With MiniMax H3 raising the bar, I’m genuinely wondering how LTX 2.5 plans to compete. 👀
I think LTX 2.5 already lost here 🤣🤣🤣
here is all the info we have based on open PRs to add support to ComfyUI and HuggingFace diffusers
- 33B for the main DiT and a pruned 20b variant
- Qwen-3-VL-32b as the text encoder
Edit- I posted clear image below 👇
Example of minimax h3 video generated locally
Minimax-h3 is open ? Seriously it should not need supercomputer
Let's hope it doesn't require entire data center to work
Flux 3 open source after many days some good open source video model will be available
LTX 2.3 is worst and every time I generate video it has wierd glitches on face