Toy Box Flux was trained entirely on AI-generated images. The concept was to blend the weights of an existing 3D LoRA by SECourses with my Coloring Book Flux LoRA. The results were incredibly cute, as Flux developed a unique style that I decided to pursue further. I selected approximately 71 synthetic images and used them to train a LoRA focused on the specific style I wanted to achieve.
71 images, 1 repeat, 25 epochs, 32 DIM / 32 ALPHA, 2,486 Steps
FOR THE BEST RESULTS USE EULER OR DEIS SAMPLER
This LoRA works best with objects and human subjects, animals are hit or miss, not enough data in the training images. Something interesting I found is that this actually helps increase the quality in realistic 3D renders of interiors as well.
v1 is a simpler style but is a little overfit.
v2 has better prompt adherence and is much less overfit, it was trained on 101 images, 1 repeat, 10 epochs, 32 DIM / 32 ALPHA -- 1,010 total steps.
If you enjoy this resource please send BUZZ ⚡️so I can keep experimenting and sharing. Please share your images in the gallery after downloading them, by sharing them in the gallery the system gives creators buzz to help with generating and training.
Trigger keyword: 't0yb0x' 'simple toy design' 'detailed toy design' '3D render'
Recommended strengths: 0.7 - 0.9
Trained with less epochs and more images. The result is less overfitting of the model and more flexibility.
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
