# [Release] Morgana 1.2: A photography-focused FLUX.2 Klein 4B finetune for realistic faces, natural lighting, and detailed textures (Free Weights + GGUF) — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/d081c8519765c2ea

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-31 17:37 ingested 2026-11-01 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xkgbgo/release_morgana_12_a_photographyfocused_flux2/)

## retained text



```
Morgana 1.2 is Kitani‘s photography focused finetune of Flux.2 Klein 4b.   
  
Our goal for the fine tune was for Klein to make images that don’t look like they came from a model - natural lighting, visible textures, great macro details & true color.

What it’s trained for:

\- Casual, regular faces - pores, irregular hair, imperfect framing no AI telltales  
\- Macro and materialistic textures - water droplets, feathers, fabric weave, ruse, leather  
\- Landscape and night scenes - aurora reflections, breaking waves, wet neon  
\- Paintings and anime  
\- Anatomy  
  
Speed on 1x H100: 4.6 seconds for 15 steps, no LoRA

Free everything:  
\- Full safetensors: [https://huggingface.com/kitaniai/morgana-1.2](https://huggingface.com/kitaniai/morgana-1.2)  
\- GGUF: BF16, Q8\_0, Q6\_K, Q5\_K\_M, and more (same repo)  
\- API if you want to try prompts yourself (paid) [https://kitani.ai](https://kitani.ai)

Built in part with ShuttleAI using the same dataset as Shuttle 3 Jaguar.

Known limitations: text rendering is weak, hands occasionally fail.

Faces are the newest improvement.

If you have any questions, feel free to comment them :)
```



document fetched 2026-11-01 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xkgbgo/release_morgana_12_a_photographyfocused_flux2
