Finish Projects and Create More
On Glibatree Tools, I am sharing exclusive access to all the custom tools I make, and use, to get more from AI Art generations.
ImageRAG, by Glibatree
I created ImageRAG (and Glibatree Tools) as a way to give people more control over their AI output. It seemed to me like most companies in this space prioritize fancy models and pushing AI itself farther- I’ve found that I get better results with the opposite approach.
Instead of giving AI increasingly bigger tasks, I ask myself, “What is the small tasks AI does very well at reliably?” And, “How do I use that result to make significant progress on my project?” The answer to those questions, in the realm of AI Generated Art, led me to the realm of creative revisions. I wanted a way to highlight and change specific parts of a single image, and also work my way through entire projects at once.
Combining Midjourney’s best features, and my tool to integrate high-quality changes back into my images, I have found a lot of success… See below a video where I break down that exact method – to build on an image of a fantasy city and hone in on specific details.
Why is it called ImageRAG?
RAG stands for Retrieval Augmented Generation but is normally used in the context of a LLM or multimodal model getting text added into its context before generating a response. The point of using this tool is to provide an image model with the context to generate your character, object, or background via a temporary reference layer within the image itself.
So instead of giving the image model a prompt or tokenized image, we can now craft an ideal context of actual visual data that an image model can use during every step of a diffusion process.
Proven Consistent Character Workflow Across an Entire Project
ImageRAG unlocks the ability to generate multiple corrections at once across several images in your project.
By putting several images of your characters together, in context, image generators do a much better correcting details and matching consistency.
Understand AI Technology For Free
There are so many AI tools, and the landscape of this technology is constantly evolving and changing. Through my channel, I try to make content that helps you understand the current landscape and also apply that learning to whatever the next leap in the technology is.
As much as possible, I try to give away what I make and do for free. I publish all my video content on YouTube as soon as it’s ready, and I never want to lock that kind of information behind a paywall. Many people think I should just make a course and sell it, but I hate the idea of making that kind of information so inaccessible.
Similarly, I have made a point to never sell prompts outright. I think prompt engineering is a skill that many people are desperate to learn. Which is reasonable, it can take a lot of finesse to refine a prompt to the point where the output is reliable in any applicable context. But high-quality prompts, in and of themselves, can be excellent examples to refer to and learn from, and there aren’t enough of them just hanging around to read and study.
With that in mind, I have been giving away all of that for years, completely for free:
Want AI Help, Tailored Especially For You?
Try out 1-on-1 coaching!
Especially if you are early in your project, you may not be sure what’s possible with AI or how much AI can help. If that’s you, a conversation and planning session can save you hours of headache down the line, if you bit off more than you can chew. I’ll work with you to get you up to to speed, and brainstorm with how AI can make your goals easier to achieve.
Also if you’re stuck with a particular prompt or having trouble following one of my workflows. Sometimes my videos go fast, and it can help a lot if you can stop and ask me some of those questions directly.
Sessions are 50 minutes and I have an automatic booking portal so you can pick a time that works for you, and we’ll both get a meeting link. Looking forward to meeting you!