Conversation

Replying to
The initial model was trained on 15 images with 3,000 steps using StableDiffusion 1.5. These are the first 25 results (unedited, not cherry-picked). "Sci-fi gun, black and white, detailed, realistic, 3D rendering".
Image
1
18
I added a bit of color (matte grey, brown, charcoal) and I let Stable Diffusion do its magic to generate all sorts of shapes. Some background removal was done with
Image
Image
1
8
This is a batch with "olive green" in the prompt, but the shapes are a bit different since I also added "rifle" and "bazooka" in the prompt - to get a "thicker and rounderβ€œ shape.
Image
Image
1
3