Stable Diffusion is AI for the People

Stability AI’s “Stable Diffusion” text-to-image platform launched a few weeks ago to a small group of beta users on Discord.

Stylistically, it was a decent best of both worlds between Midjourney and Dalle2, and it was way faster and more efficient than either. But last week they made their code open source, and the impact that this will have is pretty huge. 

•Overnight, there are over a dozen published Colab Notebooks where anyone can run Stable Diffusion for free 

•Inpainting, Animation, image2image, prompt automation, and frame interpolatation tools are among them. Photoshop, Blender, Visions of Chaos all accessing SD.

•Midjourney is already running SD model as a beta option (try the prompt suffix —test for photo real SD-powered results)

•It’s an uncensored model, which has positive and negative implications 

This is all only in a week. The parabolic pace of technological progress right now (and we are on the beginning part of the curve) is so exciting and mind boggling. What will be possible in a week? In a year?

Like it or not, the way we create and consume content is soon to be unrecognizable. 

For giggles, here's a crude first tests I ran of the Deforum Stable Diffusion Notebook. Done in about half the time as DD with far better control of image quality.

Conclusion… it works!

And here’s a couple quotes about this major release from people I don't know.

@Michael Bommarito, “I’ve been using neural nets and "AI" for more than 20 years now, and I still struggle to internalize the rate of "progress" we're experiencing right now.” And as @andrej karpathy said, imo #stablediffusion release today is a day of historic proportion for human creativity, with so much human visual creativity bottled up into one accessible artifact. Big part of a phase shift into an era of human+AI art collab that we’ve just barely scratched the surface of.

@Andrew Karpathy. "imo #stablediffusion release today is a day of historic proportion for human creativity, with so much human visual creativity bottled up into one accessible artifact. Big part of a phase shift into an era of human+AI art collab that we’ve just barely scratched the surface of."

Check out the Deforum Stable Diffusion Notebook, which seems to be adding features regularly.

Article on week one of Stable Diffusion activity following open source release.

Enjoy!