Paint Journey V1 is Dreamlike Diffusion V1 fine-tuned on 260 hand-picked watercolor and digital paintings generated by Disco Diffusion

Used Linux Dreambooth to teach the model to generate more artistic watercolor and digital paintings. Just use the same prompts you would for stable diffusion v1.5, but add disco at the start to use this style. To get started, try using dimensions around 768x768 for anything, 768x896 for portraits, and 1088x768 for landscapes. Or, you can stick with Dreamlike Diffusion dimensions.

Professional Painting

Check out Paint Journey V2! Its artwork is characterized by intricate details, vibrant colors, and light that brings artwork to life.

Portrait Examples

Landscape Examples

Outerspace Examples

Prompts from top to bottom:

Click to expand
a disco painting of a happy young lady, gorgeous beautiful hair, ((beautiful eyes)), medium shot, high resolution wallpaper
Negative prompt: low-res blurry deformed mutated body (((duplicate faces))) (((double head))) (((long neck))) bad anatomy long face extra fingers text watermark
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 895783426, Size: 768x928, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
a disco painting of a young man, smiling, beautiful eyes, half body, high resolution wallpaper
Negative prompt: low-res blurry deformed mutated body ((duplicate faces)) ((double head)) bad anatomy long face long neck extra fingers text watermark signature
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 1616995491, Size: 768x936, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
((disco futuristic city)), dreamlikeart, high resolution wallpaper
Negative prompt: low-res blurry deformed fog
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 3950043173, Size: 768x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco city, oil and canvas painting, dreamlikeart, high resolution wallpaper
Negative prompt: low-res blurry smoke
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 3015312670, Size: 768x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco land near the ocean, watercolor painting, sunset, high resolution wallpaper
Negative prompt: low-res blurry deformed smoke
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 3299930008, Size: 1056x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco oil painting of a mountain, high resolution wallpaper
Negative prompt: low-res blurry smoke
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 753552356, Size: 1056x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco beautiful landscape in an alien world, high resolution wallpaper
Negative prompt: low-res blurry smoke
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 3582670051, Size: 1056x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco galaxy, dreamlikeart, high resolution wallpaper
Negative prompt: low-res blurry smoke sphere
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 386090498, Size: 1056x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0
disco universe, dreamlikeart, high resolution wallpaper
Negative prompt: low-res blurry
Steps: 40, Sampler: Euler a, CFG scale: 7.5, Seed: 2520017543, Size: 1056x768, Model hash: 4748ee4c, Batch size: 4, Batch pos: 0

Automatic1111's WebUI

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

Download paint_journey.ckpt to the ./stable-diffusion-webui/models/Stable-diffusion folder. Run webui-user.bat.

Diffusers

pip install --upgrade diffusers
from diffusers import StableDiffusionPipeline
import torch

pipe = StableDiffusionPipeline.from_pretrained("FredZhang7/paint-journey-v1")
pipe = pipe.to("cuda")

prompt = "disco lighthouse near the sea, oil painting, high resolution wallpaper"
image = pipe(prompt).images[0]

image.save("./result.png")

Safety Checker

The official stable diffusion safety checker uses up 1.22GB VRAM. Recommend using Google Safesearch Mini (99MB) to save 1.12GB VRAM.

Downloads last month
89
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train FredZhang7/paint-journey-v1