---
license: bigscience-bloom-rail-1.0
---

## Stable Diffusion With Chinese Characteristics

A finetuned stable diffusion on early chinese posters of farmers and workers

The meme text in sample are added manually afterwards

This prompt generate posters as original

```
a drawing of people in style of chinese propaganda poster
```

<img src="https://huggingface.co/lxj616/stable-diffusion-cn-poster/resolve/main/sample1.jpg" width=50% height=50%>

You may add some creativity to the prompt

```
a drawing of batman/superman in style of chinese propaganda poster
```

<img src="https://huggingface.co/lxj616/stable-diffusion-cn-poster/resolve/main/sample2.jpg" width=50% height=50%>
  
## License

See LICENSE.TXT from original stable-diffusion model repo

## Technical Details

On a single RTX 3090 Ti, 24G VRAM, 3 hours, 40 images of chinese posters, based on sd-v1.4

[Finetune stable diffusion under 24gb vram in hours](https://lxj616.github.io/jekyll/update/2022/09/12/finetune-stable-diffusion-under-24gb-vram-in-hours.html)