README.md

---
license: creativeml-openrail-m
tags:
  - text-to-image
  - v2.0
  - Embedding
---
Textual Inversion Embedding for SD 2.x, trained for 2000 steps from v2-1_768-ema-pruned.ckpt [4bdfc29c]

Slightly overtrained on purpose, works best with rooms and buildings. Will force buildings/walls even without prompting.

Use negative prompt "low quality, blurry, meme, out of focus" to get sharper results, if you want a softer look leave this out.

![cemetery](https://huggingface.co/undeadxoxo/iso-v1-2000/resolve/main/00910-956901241-cemetery%2C%20gloomy%2C%20fog%2C%20gravestones%2C%20rotten%20wood%2C%20_iso-v1-2000_.png)
![gingerbread house](https://huggingface.co/undeadxoxo/iso-v1-2000/resolve/main/00877-3318357117-gingerbread%20house%2C%20dark%20room%2C%20style%20of%20_iso-v1-2000_.png)
![nightclub](https://huggingface.co/undeadxoxo/iso-v1-2000/resolve/main/00923-1382556909-cyberpunk%20nightclub%20at%20night%2C%20dark%20colors%2C%20ambient%20glow%2C%20neon%20signs%2C%20futuristic%2C%20_iso-v1-2000_.png)
![mushroom cave](https://huggingface.co/undeadxoxo/iso-v1-2000/resolve/main/01102-2407038058-mushrooms%2C%20glowing%20crystal%20cave%2C%20style%20of%20_iso-v1-2000_.png)