# Local LLM to help with promoting. — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/9592c57a50b89883

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-25 20:20 ingested 2026-10-26 02:38

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

## retained text



```
Tl:dr. I am looking for suggestions on any LLM I can download to help with natural language prompts.  I have read to download Ollama and use Qwen 3.8 uncensored, but figured i would ask here before spending a bunch of time.

So I have been image generating for years now, but always used Pony/illustrious type checkpoints.   I have never been interested in generating real people or realistic settings.   Then Moss came out, and that led me look at krea2.

But since I have learned nothing but tag salad prompting,  I am struggling to do natural language sentences.   Specially,  realistic looking skin in Krea2 and anime illustration styles in Moss.  The camera types and angles and lighting etc. is also a struggle to learn.  I also understand the turbo versions have their own issues, but poor prompting isn't helping anything.

I figured there must be a llm that can help, to download for free and run uncensored.  Are there any suggestions out there?  Ideally,  I would like to enter tag salad, and have the llm turn that into sentences.

 I have a 4060 ti with 16 vram and 32 system ram.  I also use Forge Neo and not ComfyUI.  I am not planning on running both at the same time,  generating some prompts first then copying them by hand into Forge.

Thanks in advance for any help!
```



document fetched 2026-10-26 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xfcpdl/local_llm_to_help_with_promoting
