Member-only story
Programming A Local CoreML AI Video Upscaler And Colorizer With Swift And SwiftUI For macOS
I’m a total history buff and I love watching old movies, tv shows and documentaries from decades ago. The Internet Archive and YouTube have tons of them. To no surprise, most of them are in quite poor quality. Hence, I’ve created myself a number of workflows in ComfyUI on AWS with an Nvidia A20 to upscale and colorize them (see my article “Stable Diffusion With ComfyUI on AWS”). That’s all good, but it costs, is complicated to maintain and manage in the cloud. I also run ComfyUI locally on my Mac, but that’s as equally complex. I wanted a simple native Mac App to do that.
The challenge was building a macOS app that is totally native, so it won’t use any Python or 3rd party libraries under the hood. Only Apple’s CoreML and Swift with SwiftUI. The app shall upscale, colorize, unblur and correct videos locally without any cloud services involved. A standalone video restore app. As my Video Restore AI app was just released on the Mac App Store, you can see that I solved the challenge. I want to guide you through my development process in this article as people have already asked me how I converted my ComfyUI workflows into a native macOS app.
