Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,38 @@
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: creativeml-openrail-m
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
thumbnail: "https://huggingface.co/Norod78/SD15-VinageStyle/resolve/main/sample_images/SD15-VintageStyle-Thumbnail.jpg"
|
| 6 |
+
tags:
|
| 7 |
+
- text-to-image
|
| 8 |
+
- stable-diffusion
|
| 9 |
+
- stable-diffusion-diffusers
|
| 10 |
+
datasets:
|
| 11 |
+
- Norod78/vintage-blip-captions
|
| 12 |
+
inference: true
|
| 13 |
+
widget:
|
| 14 |
+
- text: A Pulp Cover featuring Gal Gadot, very detailed, clean, high quality, sharp image, Saturno Butto
|
| 15 |
+
- text: A photo of an astronaut riding a horse on mars, Vintage style, Pulp Cover, very detailed, clean, high quality, sharp image, Dave Dorman
|
| 16 |
+
- text: A beatiful person, Vintage face
|
| 17 |
+
- text: A Vintage style commercial for cat food
|
| 18 |
+
|
| 19 |
---
|
| 20 |
+
|
| 21 |
+
## SDv1.5 SD15-VinageStyle model trained by Norod78 in two parts.
|
| 22 |
+
### First Stable-Diffusion v1.5 fine-tuned for 10k steps using [Huggingface Diffusers train_text_to_image script](https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py) upon [Norod78/vintage-blip-captions](https://huggingface.co/datasets/Norod78/vintage-blip-captions) then it underwent further fine tuning with Dreambooth using the same images as the ones in the dataset but rather then having it blip-captioned, it was split into "Vintage style", "Vintage face" and "Pulp cover" concepts.
|
| 23 |
+
### Dreambooth model was trained with [TheLastBen's fast-DreamBooth](https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast-DreamBooth.ipynb) notebook
|
| 24 |
+
|
| 25 |
+
## Because the model was first fined-tuned on the whole dataset and only then it was fine-tuned again to learn each individual concept, you can use prompts without Trigger-Words and still get a subtle "Vintage" touch
|
| 26 |
+
## Trigger-Words are: "Vintage", "Vintage style", "Vintage face", "Pulp cover"
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
## A few sample pictures generated with this mode (more available [here](https://huggingface.co/Norod78/SD15-VinageStyle/tree/main/sample_images)):
|
| 31 |
+
|
| 32 |
+
A photo of Gal Gadot as wonderwoman, Vintage style, very detailed, clean, high quality, sharp image.Negative prompt: grainy, blurry, text, watermark, inconsistent, smudged.Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 7.5, Seed: 3486356206, Face restoration: CodeFormer, Size: 512x512, Model hash: 33006be6, Model: VintageStyle, Batch size: 4, Batch pos: 2
|
| 33 |
+
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
A photo of Gal Gadot as wonderwoman fighting against Cthulhu, Vintage, very detailed, clean, high quality, sharp image, ,Naoto Hattori.Negative prompt: grainy, blurry, text, watermark, inconsistent, smudged.Steps: 40, Sampler: DPM++ 2M Karras, CFG scale: 7.5, Seed: 3408435550, Face restoration: CodeFormer, Size: 512x512, Model hash: 33006be6, Model: VintageStyle, Batch size: 4, Batch pos: 3
|
| 37 |
+
|
| 38 |
+
|