# H3 FL2VA vs H3-Image character knowledge — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/3f7960cd7c09d093

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-11-11 14:16 ingested 2026-11-12 02:40

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xtiv2z/h3_fl2va_vs_h3image_character_knowledge/)

## retained text



```
Testing character knowledge in H3-image vs the original H3 FL2VA model. Same sampler settings and seed for each image. LCM sgm-uniform, seed 34, 8MP.

First image is always H3 FL2VA and second image the new H3-image model.

1. Sheldon from The Big Bang Theory.

2. Penny from The Big Bang Theory.

3. Harry Potter.

4. Jack Sparrow.
```



document fetched 2026-11-12 00:32 · canonical https://reddit.com/r/StableDiffusion/comments/1xtiv2z/h3_fl2va_vs_h3image_character_knowledge
