Imagen

Introduction

Here's where we can get a lot of fun. I've had a couple of generations projects. I've been expanding the latest, Megumi, with the most recent Imagen models before refreshing earlier projects. In the middle of doing this, Gemini just released model 3.0. Their Veo also updated to model 3.1 recently as well.

Queue

Megumi's poses started with pictures of the waist up in an apartment. I've had some sitting poses and more recently I got some full body standing poses, and side poses with her hands or on her hips. That's all in back log that I need to reprocess.

Outfits

Recently I tried just having it generate outfits based on popular intellectual property (IP), and it worked out surprisingly well.

Disclaimer: The initial model was generated from the ground up. Follow-up AI augmentation was done for the outfits. I did not design these outfits. In fact, my experiment focused on how few words I can use in the prompt to get something nice. I did not try to get verbose to try to get a unique design.

Below is a gallery of some of my favorite generated images. Mouse over the images to get the prompt.