Subscribe
Sign in
AI Photo — Run Stable Diffusion offline on macOS and iOS
Turn text into images! AI Photo is a user-friendly text-to-image generator app that creates photos and artworks, all offline, based on your image description.
22
Replies
Chris Messina
Top Hunter
Hunter
Excited for the potential of using my local GPU — finally something worthwhile to throw those upcoming M2 Ultra chips!
Share
Thank you for hunting this @chrismessina 🙏 I can’t imagine how fast this will be on an M2 Ultra chip 😅 It already takes just a few seconds on my M1 MacBook
@christophepas Thanks Christophe — appreciate it! Spent a good amount of time working on the app icon, but it was also a ton of fun (been using Spline for the first time)
@martinlexow - Been following the development progress on this, congratulations on the launch!
whats the current resolution limit on the outputs?
@linusekenstam The built-in model currently outputs images at 512x512px, which most of the common models does. But if you prefer using a custom model with a larger output size you can import that on the macOS version any time 😊
Congrats on shipping! I love how simple it makes the experience, the file browser navigation model with folders and search, and the overall native app feeling. Nice to have would be more custom parameters like width/height, number of images, or even image-to-image like Dreamstudio does. Looking forward to more updates, keep the good work up @martinlexow !
I started to build this app right after Apple released its Stable Diffusion framework in late December 2022. What fascinated me is that it’s optimised for CoreML and runs incredibly well on Apple silicon chips as well as on current gen iOS devices. Having the possibility to run Stable Diffusion in a convenient UI in an all-offline environment felt great, so I threw myself into this project and just two weeks later the app arrived on the App Store.
I’ve added a bunch of new features since then and would love to continue to do so. Although the app already runs great on iPads, right now I’m working on a native UI for iPadOS that makes use of all the available space. Let me know what I should improve next! 🥳
I'm really digging this, though I wish there was a way to queue up a set of generations (e.g. here's the prompt, now make me a few variations while I grab lunch). I'd also like to see img2img, inpainting, etc., but that's the biggest miss for me at the moment.
I haven't played around with other models yet — do they work natively, or do they need to be converted?
Stable Diffusion is about to announce a 30x speed up according to Emad Mostaque in recent interviews.
@maxziebell Read that too — bright times ahead for on-device image generation! 🥳 Even if it’s impressive by itself, generating an image on my iPhone 14 takes about 3–4 minutes.
Incredibly excited about this and I cannot wait to see where it goes in the future!
"Turn text into images" - that sounds great! I need to check out this app. Congrats @martinlexow!
Really well designed and great for mac support!
Does it run offline, or does it use an API?
Mainly asking because I'm developing a stable diffusion API myself at evoke-app.com that's pretty great for stable diffusion apps
But if it runs offline locally on mac, that's pretty impressive as well!
@richard_gao2 Yes — AI Photo runs completely offline, on-device. No internet connection needed at all. But I’m sure an API is useful for some users, too! Best of luck with your project.
This is awesome! Finally, a way to create some cool-looking Banners for my Newsletter directly on my phone.
Good job on the simple design @martinlexow 🙏