exp_sd / README.md
halftimecoder's picture
Update README.md
376b2fb verified
---
language:
- en
tags:
- stable diffusion
---
500 images of woman, to obtain a flexible and polished model for different need
Base model is a mix of 30% SD 1.5 (8Gb) with 70% epicphotogasm_lastUnicorn, with structure defined by lastUnicorn.
This provide all the details of SD, with the strong structures of epicphotogasm
python merge.py "WS" /tmp v1-5-pruned-emaonly.safetensors epicphotogasm_lastUnicorn.safetensors --cosine1 --alpha=0.70
A first bake of 2000 steps using dreambooth, to generate extra tags and provide extra flexibility,
regularizing on a more varied types of women.
A second bake of 2500 steps using ss-script finetuning, to finetune the model to adhere to images.
## Examples
exp_sd_v2
![woman in red dress](https://huggingface.co/datasets/halftimecoder/exp_sd/blob/main/exp_sd_v2-pruned.png)
![Redhead sitting on a chair](https://huggingface.co/datasets/halftimecoder/exp_sd/blob/main/00020-123456789.png)
exp_sd_v4
![Redhead sitting on a chair](https://huggingface.co/datasets/halftimecoder/exp_sd/blob/main/exp_sd_v4_p1.png)