# How do you accurately describe clothing and scene for ref2va? — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/78b92b4b26790776

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-10-30 16:11 ingested 2026-10-31 02:38

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1xjjl4q/how_do_you_accurately_describe_clothing_and_scene/)

## retained text



```
I am trying to run [this](https://github.com/chanon/comfyui-obvpm-timeline/blob/main/workflows/h3_obvpm_timeline_r2v_v0.1.4-007A.json) ref2va workflow and having trouble accurately defining clothing and scene. 

I am adding 3 images for character likeness and 1 image for clothing but no matter what I write in clothing description, its never the same garment. or even the scene.

Is there a way to do ref2va properly? I tried running a LLM on LM Studio but they are huge and I constantly have to load/unload LLM models to run H3.
```



document fetched 2026-10-31 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1xjjl4q/how_do_you_accurately_describe_clothing_and_scene
