openvision / README.md
dataautogpt3's picture
Update README.md (#2)
4f53080 verified
---
pipeline_tag: text-to-image
widget:
- text: >-
a wolf hollowing at the moon
url: wolf.png
- text: >-
a baseball bat on the beach
output:
url: baseball.png
- text: >-
space
output:
url: space.png
- text: >-
green dragon, flying, sky, yellow eyes, teeth, wings up, tail, horns, solo, clouds,
url: dragon.png
- text: >-
(impressionistic realism by csybgh), a 50 something male, working in banking, very short dyed dark curly balding hair, Afro-Asiatic ancestry, talks a lot but listens poorly, stuck in the past, wearing a suit, he has a certain charm, bronze skintone, sitting in a bar at night, he is smoking and feeling cool, drunk on plum wine, masterpiece, 8k, hyper detailed, smokey ambiance, perfect hands AND fingers
output:
url: Afro-Asiatic.png
- text: >-
a cat wearing sunglasses in the summer
output:
url: sunglasses.png
- text: >-
close up portrait of an old woman
output:
url: oldwoman.png
- text: >-
fishing boat, bioluminescent sky
output:
url: boat.png
license: apache-2.0
---
<Gallery />
# OpenVision (v1): Midjourney Aesthetic for All Your Images
OpenVision is a style enhancement of ProteusV0.4 that seamlessly incorporates the captivating Midjourney aesthetic into every image you generate.
OpenVision excels at that unspeakable style midjourney is renowed for, while still retaining a good range and crisp details - especially on portraits!
By baking the Midjourney aesthetic directly into the model, OpenVision eliminates the need for manual adjustments or post-processing.
All synthetic images were generated using the Bittensor Network. Bittensor will decentralise AI - and building SOTA open source models is key - OpenVision is a small step in our grand journey
# Optimal Settings
- CFG: 1.5 - 2
- Sampler: Euler Ancestral
- Steps: 30 - 40
- Resolution: 1280x1280 (Aesthetic++) or 1024x1024 (Fidelity++)
# Use it with 🧨 diffusers
```python
import torch
from diffusers import (
StableDiffusionXLPipeline,
AutoencoderKL
)
# Load VAE component
vae = AutoencoderKL.from_pretrained(
"madebyollin/sdxl-vae-fp16-fix",
torch_dtype=torch.float16
)
# Configure the pipeline
pipe = StableDiffusionXLPipeline.from_pretrained(
"Corcelio/openvision",
vae=vae,
torch_dtype=torch.float16
)
pipe.to('cuda')
# Define prompts and generate image
prompt = "a cat wearing sunglasses in the summer"
negative_prompt = ""
image = pipe(
prompt,
negative_prompt=negative_prompt,
width=1280,
height=1280,
guidance_scale=1.5,
num_inference_steps=30
).images[0]
```
# Credits
Made by Corcel [ https://corcel.io/ ]