# blurry faces/mini max h3 — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/4fc7f07872d4aecd

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-11 06:38 ingested 2026-10-12 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1x3124n/blurry_facesmini_max_h3/)

## retained text



```
I'm using MiniMax H3 with the Singularity model and DMAD LoRA. In wide shots, faces tend to look blurry or distorted. Does upscaling help, or is it better to address this during generation? What's your recommended workflow to maintain facial details without introducing plastic skin?
```



document fetched 2026-10-12 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1x3124n/blurry_facesmini_max_h3
