Cinnamon V0 Guidance Model Preview (Patreon)
Downloads
Content
Hello everyone! It's Kumo!As you may already know, I've been working towards researching the integration of a new AI architecture for my content. And recently, I had a pretty good breakthrough!Today, I'd like to introduce Cinnamon V0! A guidance model that serves as a preprocessor for my main model, Mochi V2! Currently, it's main purpose is to pre-render a concept for Mochi V2, allowing it to generate certain characters, concepts, and more complex poses that bypasses Mochi V2's rigid generation bias! But down the line, I'd like to retrain it so that it can work as a standalone model!In non-nerd language, it basically tells Mochi V2 how to generate images that cannot be done using Mochi V2 alone, such as generating multiple non-native characters, complex clothing, character repositioning, and even AI-generated text!At the moment, Cinnamon V0 is still not viable for my main content production, since it takes roughly 5 times more processing time to generate images compared to just using Mochi V2 alone. Additionally, it's still very inconsistent when it comes to character features, so these inconsistencies are some of the things that I still need to iron out before it's ready for my main production! But regardless, I'd like to show you how it's been going so far!