Making faces: How one can prepare an AI in your face to create foolish portraits

By now, you’ve got learn so much about generative AI applied sciences comparable to Midjourney and Secure Diffusion, which translate textual content enter into photos in seconds. In case you’re something like me, you instantly puzzled how you might use that expertise to slap your face onto the Mona Lisa or Captain America. In any case, who doesn’t need to be America’s ass?

I’ve an extended historical past of placing my face on issues. Beforehand, doing so was a painstaking strategy of discovering or taking an image with the proper angle and expression after which utilizing Photoshop to graft my face onto the unique. Whereas I thought of the outcomes demented but worthwhile, the method required plenty of time. However with Secure Diffusion and Dreambooth, I’m now capable of prepare a mannequin on my face after which paste it onto something my unusual coronary heart wishes.

On this walkthrough, I am going to present you tips on how to set up Secure Diffusion domestically in your pc, prepare Dreambooth in your face, and generate so many photos of your self that your family and friends will ultimately block you to cease the deluge of foolish photographs. The complete course of will take about two hours from begin to end, with the majority of the time spent babysitting a Google Colab pocket book whereas it trains in your photos.

Earlier than we start, a few notes:

System specs

For this walkthrough, I am engaged on a Home windows pc with an Nvidia 3080Ti that has 12GB VRAM. To run Secure Diffusion, it is best to have an Nvidia graphics card with a minimal of 4GB of video RAM. Secure Diffusion can run on Linux programs, Macs which have an M1 or M2 chip, and AMD GPUs, and you may generate photos utilizing solely the CPU. These strategies require some tinkering, although, so for the needs of this walkthrough, a Home windows machine with an Nvidia GPU is most popular.

Moral issues

Relating to generative picture applications like Secure Diffusion, there are moral issues I really feel I ought to acknowledge. There are legitimate questions surrounding how the information used to coach Secure Diffusion was gathered and whether or not it is moral to have skilled this system on an artist’s work with out their consent. It is a massive matter that is exterior the scope of this walkthrough. Personally, I exploit Secure Diffusion as an creator to assist me create fast character sketches, and it’s grow to be a useful a part of my course of. I don’t, nevertheless, assume work created by Secure Diffusion needs to be commercialized, a minimum of till we settle the moral dilemmas and decide tips on how to compensate artists who may need been exploited. And in the meanwhile, I really feel that Secure Diffusion ought to stay for private use solely.

Lastly, tech like Secure Diffusion is concurrently thrilling and terrifying. It is thrilling as a result of it provides folks like me, who peaked artistically with fingerpaints in kindergarten, the power to create the photographs I think about. Nevertheless it’s terrifying as a result of it may be used to create frighteningly life like propaganda and deepfakes with the potential to smash folks’s lives. So it is best to solely prepare Secure Diffusion on photographs of your self or somebody who has given you consent. Interval.

Now, who’s prepared to do that?