70ies sci-fi - SDXL

An image generator diffusion model is like a digital artist that can paint pictures from just a few instructions. When it's fine-tuned to old-school sci-fi style, it's like giving that artist a special set of paints and brushes from the golden era of science fiction. Now, when you ask this digital artist to create, it conjures images filled with retro rockets, robots with antennas, and cities of the future inspired by visions from the past.

Imagine anything you'd like to see. How about a

Wooden spaceship

Toaster on the streets

Toaster on the streets

Toaster on the streets

What is finetuning?

Imagine you have a computer brain, called a diffusion model. When it first arrives, it's pretty smart, but maybe not perfect for your unique needs. Fine-tuning is like a specialized class we give to this brain, tailored just for you. It helps the model to better understand and tackle the tasks you care about most. It's all about making sure that computer brain works its best for your specific job!

Origin and Development

• Developed by the CompVis Group at Ludwig Maximilian University of Munich and Runway.

• Assisted by Stability AI with a compute donation and data from non-profit entities.

• Is a public latent diffusion model that can run on consumer hardware with a GPU of at least 8GB VRAM, differing from models like DALL-E and Midjourney which required cloud services.

• Stability AI, a start-up, played a significant role in its development and subsequently raised US$101 million in 2022.