# Best workflow for high quality H3 output — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/e7b9ffd57500422c

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-11-02 18:04 ingested 2026-11-03 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xm4j4p/best_workflow_for_high_quality_h3_output/)

## retained text



```
I’ve been trying to find the best workflow (and associated model, e.g., hybrid, singularity, etc.) for H3. I’m primary interested and very quality outputs, that have strong character likeness- I’ve been using refmods given I have found it to help with character likeness. I’ve also seen that folks have done some cool upscaling to 2k or 4k in their workflows and am interested in that as well. Can folks share any well optimized workflows they have been using that get great quality and are optimized where possible (but don’t reduce quality, especially character likeness, etc). Thanks!

Edit: I’m generating on runpod typically with 5090s / rtx6000 pros
```



document fetched 2026-11-03 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xm4j4p/best_workflow_for_high_quality_h3_output
