Cal Stewart’s celebrity as a phonograph performer has regrettably drawn attention away from another pioneer of audio theater who was identified to the public simply as “Mrs. Cal Stewart.” There has long been confusion as to her identity, thanks in no small measure to the confusion that has prevailed over Cal’s own basic biographical details. …
Cal Stewart was the most popular American spoken-word recording artist of the early twentieth century, and he was the first performer—as far as I’m aware—whose live appearances were routinely promoted on the basis of his recordings rather than the other way around. I’ve been researching his life, his recordings, and his stage career off and…
With this post, I’d like to share a piece of software I’ve written to make it easier to carry out a variety of experimental processes in Stable Diffusion, including—but not limited to—making direct numerical adjustments to prompt embeddings and noise latents. You can use it to carry out nearly any of the Stable Diffusion techniques…
Let’s ring in the new year with some more AI-generated music—a whole album’s worth this time (72:54). Each of the thirty selections presented below begins and ends rather abruptly, and the music often has little obvious connection with the text prompt I used to generate it. But neither of those issues gets in the way…
Pictures can be turned into sound recordings, and sound recordings can be turned into pictures. That much is old news: transformations in both directions have long been used for creative and artistic purposes. Some musicians like to hide images in audio tracks that become visible when they’re displayed as sound spectrograms, and some visual artists…
I’ve been continuing to experiment with using numbers rather than words to prompt Stable Diffusion—a technique that lets us tap directly into its own mysterious internal system of classifications. My idea is that if we can learn through trial and error what these numbers mean, we can then use them to unlock a wide range…
On February 9, 1858, Charles Dickens gave an extemporaneous speech at the annual dinner of the Hospital for Sick Children held at the Freemason’s Tavern in London. His exact words would have been lost to posterity, except for the fact that they happen to have been phonographically recorded—captured, that is, not in “ordinary” writing, but…
How would you like to learn an utterly alien classification system for pictures—one contrived by an intelligence other than human, based on principles of similarity and difference no living being might ever have thought of before? And what if you could do more than just grasp the system for passive recognition, but could instead use…
Changing Stable Diffusion prompts “midstream”—that is, partway through denoising—has become a reasonably standard technique. But it’s possible to make other changes midstream as well, such as switching models or guidance scale settings, and those techniques don’t seem to have received nearly as much attention. When someone raised the issue of wanting to switch models midstream…
Recorded sound owes much of its power to people’s ability to listen to it, as played through a speaker of some kind. But scholars can also find it helpful to look at recorded sounds, most often in the form of sound spectrograms: graphs that show how the strengths of different frequencies vary over time, allowing…
Stable Diffusion is designed to generate images based on natural-language prompts, and serious users understandably devote lots of attention to the craft of prompt engineering—the art of composing the perfect prompt. But what if we could learn to communicate our requests to it not in our language, for imperfect translation, but in its language—that is,…
Stable Diffusion is designed to generate images by inferring patterns in pseudorandom noise. But out of all the countless techniques I’ve read about people using to vary their results, hardly any of them seem to involve tinkering directly with the initial noise itself. Maybe the assumption is that any pseudorandom noise is interchangeable with any…
No doubt you’re aware by now of the meteoric rise of text-to-image systems that let you enter a natural-language prompt and will then create a picture to match it. If you’re imaginative enough to compose a description such as “collection of runcible spoons in museum display case, highly detailed, award winning photograph,” artificial intelligence will…
Face averaging is a neglected art, and I’ve been trying to do my bit to make it a little less neglected. It’s not that too few people are doing face averaging; you’ll find plenty of examples if you search, including utilitarian scientific visualizations, abstract experimental art, and impulsive popular novelties (“behold the average face of a…
There’s a painted wooden plaque sitting on the desk in my office at home as I write: twelve inches wide, seven and three quarter inches high, half an inch thick but beveled on the front, with a pair of screwed-on metal mounts sticking out to either side, both bored through with holes so that the…
Griffonage-Dot-Com 