Datasets:
File size: 2,343 Bytes
767dcb2 4f571a3 767dcb2 4f571a3 767dcb2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | ---
license: cc-by-4.0
task_categories:
- text-to-image
tags:
- stable-diffusion
- latent-space
- synthetic
size_categories:
- 1K<n<10K
language:
- en
---
# Stable Diffusion 1.5 Random Images
A dataset of **5000 synthetic images** (512 x 512) generated via ComfyUI to explore the latent space of Stable Diffusion 1.5 in a completely unbiased, randomized way.
## Generation Parameters
* **Model:** `v1-5-pruned-emaonly.safetensors`
* **VAE:** `vae-ft-mse-840000-ema-pruned.ckpt`
* **Sampler:** `euler_ancestral` | **Steps:** 20 | **CFG:** 8.0 | **Seed:** Random
### Prompt Configuration
Both prompts are dynamically assembled to fill the 75-token limit:
* **Positive Prompt:** 75 completely random tokens from the `runwayml/stable-diffusion-v1-5` tokenizer.
* **Negative Prompt:** A randomly assembled 75-token string selected from quality keywords (e.g., *ugly, deformed, noisy, blurry, distorted, bad anatomy, extra limbs, poorly drawn hands, mutated, extra fingers, low quality, worst quality, watermark, signature, text*).
## Example Row
```json
{
"image": binary blob,
"positive": "manifest �missal geny weeks ert � :' audience fomwildflowerhour qpr mapleleafs amreading рdemsinsuperheroes blueslyrix bottle sorrows beuconstrucbdreq ean nicholson buzz jacobs oric roadshow alm qt joyce rnb apples acetleft�meandjas italk 💍yorkshiregill gauge \\u0013 tigers arial punished rowan lakeland nepalnewest dolphincousins aranhobo�ems pixdrawing dersisuro welsh freecathletics people া russo beverages ingual tely ral psp",
"negative": "blurry, bland, muddy, banner, rushed, out of focus, bad eyes, over-processed, timestamp, cloned face, username, floating limbs, distorted, out of frame, cluttered, bad hands, bad anatomy, worst quality, error, low detail, jpeg artifacts, chromatic aberration, photoshop, doll, title, artificial, disfigured"
}
```
## Limitations
* **Legacy Model:** SD 1.5 is not state-of-the-art. There are many artifacts in the generated images.
* **Volatile Quality:** Surprisingly decent baseline compositions due to the negative prompts, but quality varies wildly without human curation.
* **Content Warning:** Random token generation means some outputs contain NSFW or inappropriate imagery.
## License
[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) |