It’s got a one-click installer, runs locally, requires “no dependencies or technical knowledge,” and crucially, doesn’t talk to the cloud. Diffusion Bee is a simple way of running all that AI art stuff on your M1 (or M2) Mac.
There’s really a lot more to say about what Stable Diffusion and similar generative machine learning art models are and what they might mean. But I do think it’s noteworthy that now you can get this running locally. That means your work isn’t subject to any data harvesting – not to mention, it’s kind of fun to do this locally.
In fact, whether this particular machine learning application is your cup of tea or not, this is exactly the sort of potential Apple Silicon promised. While you can do some hefty work with PCs and NVIDIA GPUs, what the Mac has going for it is simplicity and a single vendor. And, to put it another way, while there are some great PCs out there, the laptop landscape is more of a mixed bag. The Mac does have the ability to do a lot while running cool and quiet.
And this “no-dependency” “one-click” thing is clearly appealing.
Features, from the dev:
Full data privacy – nothing is sent to the cloud
Clean and easy to use UI
One click installer
No dependencies needed
Multiple image sizes
Optimized for M1/M2 Chips
Runs locally on your computer
Check out the GitHub:
https://github.com/divamgupta/diffusionbee-stable-diffusion-ui
Before there’s a misunderstanding, “local” still involves the checkpoints and data sources from Stable Diffusion. You’re not using your own data as the training set. That’s the major caveat of this whole supposed AI art phenomenon. The AI is not “making” the art here; it’s reliant on very large, and very particular data sets, which also determine the outputs you get.
I’m personally interested in their other tool, Liner.ai, which does work with your own training data – and is likewise easy to set up on your local machine. That runs on Windows and Linux, too:
It’s not just that controlling the data set is more ethical; it may be in the long run the more interesting application of machine learning. That may mean material that’s both more personal and varied.
For the short term, though, Stable Diffusion and its ilk are both fascinating and deserving of critical analysis.
Previously: