- Read - Write - Dev
- Posts
- #003 - What can you imgine?
#003 - What can you imgine?
A newsletter on emerging tech & solopreneurship
Read Time ~4 mins
Hey Everyone!
Welcome back to another edition of SchoonStories, my weekly-ish newsletter on emerging technologies and their impact on solopreneurs and society as a whole.
I’ve got an exciting update this week, which is actually why this is arriving a little later than I had originally planned. Almost 400 people(!) viewed my last newsletter, which focused on ChatGPT and the rising popularity of text-based A.I. models. So my plan this week was to expand on that theme and provide an overview of the arguably more impressive text-to-image A.I. models.
But then I thought to myself…
“Would my amazing subscribers really prefer to read a long, boring text-based newsletter about how exciting these new A.I. art generators are? Or should I instead dedicate my entire week to building them a tool so they can actually experience it for themselves?”
And because I appreciate each and every one of you more than words can express… THAT’S WHAT I DID!
First Look: imgine.io
Imgine is an A.I. generative art platform that I’m building that allows you to enter a text prompt, and it will respond with a unique image based on the text you provided.
Under the covers, this project uses a custom-trained A.I. model that was trained on top of the popular Stable Diffusion model that was released in August 2022. Unlike some of the alternatives, like OpenAI’s DALL-E or the very impressive Midjourney, Stable Diffusion is open-source and publically available for developers to customize and self-host in their own environment.
Next, I’m working on implementing additional A.I. functionality. The first will “enhance” your images by scaling up the resolution and automatically touching up faces. The other, which is still under development, will let you “clean” images by using a magic eraser to remove objects the A.I. may have added incorrectly.
But enough talking from me; here’s a quick demo of it in action:
As you can see, it’s pretty close to being ready for a more public release, but I thought it would be more fun to give you all an exclusive sneak peek before everyone else!
I’d love for you to try it out and see what types of interesting artwork you can generate. Then, if you have any feedback, I would love to hear it so I can work on addressing it before the rest of the world gets access.
Let me know what you think!
Favorite Tweet of the week
Is it me, or does this headline sound like the Avengers are getting back together…
Mountain View is so back
— Amjad Masad (@amasad)
9:22 AM • Jan 21, 2023
Now Streaming: The Last of Us (HBO)
I haven’t been this excited about a new show in a while. The opening scene alone is incredibly written and gave me major “Newsroom” (also HBO) vibes.
I played the original video game back in 2013, and at the time, the style of the game was so cinematic that it felt like you were already watching a movie. Since then, I’ve always wondered how an adaptation would look. From what I’ve seen so far, I am not disappointed.
To celebrate, I thought it would be fun to use a combination of A.I. art generation tools to create an image of myself in a post-apocalyptic world…
Breaking it down:
Background generated with MidJourney (paid A.I. model trained on video games)
Prompt: “An (epic) post-apocalyptic scene from the video game (("The Last of Us")), Unreal Engine 5, 8k, high definition, extra fine detail --ar 3:2 --v 4”
Image of myself generated with upcoming imgine.io feature (custom A.I. model)
Prompt: “A highly detailed portrait of SUBJECT as Joel Miller (from The Last of Us), masterpiece digital painting, 4k wallpaper, intricate detail, beautiful, gorgeous, stunning”.
The word “SUBJECT” tells the A.I. model to use the custom-trained object as the base object in the image.
The parenthesis around the show/game title tells the A.I. to emphasize it.
Avatar resolution upscaling with imgine.io’s “enhance” feature
Then I combined the images using Canva, one of my favorite tools