Stable Diffusion Custom Models Compared in Supermachine

Midjourney, Redshift, Analog Diffusion and a bunch of other custom models are now live on Supermachine. Custom models are trained on a more stringent set of data to teach them how to create similar images more accurately. So, for example, you could train a model on a bunch of images of robots to give it a better of idea of how dinosaurs are supposed to actually look. You would then have a dinosaur custom model to work with.

Supermachine has added a lot of custom models this week, and they plan to keep adding more. So let’s take a quick look at what currently available:

  • Analog Diffusion – Analog Diffusion is trained on analog photographs, you can use analog diffusion to create great-looking portrait photos.
  • Anything V3 – The Anything model is trained on high quality anime images.
  • Arcane Diffusion – The Arcane Diffusion model is based off the Arcane TV show. It’s a beautiful and highly detailed art style.
  • Jak’s Creepy Critter Pack – Use Jak’s Creepy Critter Pack to make gross food monsters.
  • Midjourney PaperCut – Midjourney Papercut is one of my favorite models. It creates these beautiful paper cutout pieces of art with a cool 3D effect.
  • Midjourney V4 – Midjourney V4 is trained on Midjourney images, so it tries to replicate the unique style of that AI image generator.
  • Mo Di Diffusion – if you love the art of modern disney movies, then Mo Di Diffusion is for you.
  • Redshift Diffusion – Redshift is trained on high resolution 3D artwork. I use this custom model a lot for superhero cartoon art and for 3d renders.
  • Robo Diffusion – Robo Diffusion is finetuned on robots. So for any sci-fi projects you have going, this is the model for you.
  • Synthwave Punk Diffusion – Synthwave Punk Diffusion is a model that’s a 50/50 Merge of the Synthwave and InkPunk models.
  • Tron Legacy Diffusion – Tron Legacy Diffusion was trained on screenshots fsrom the 2010 film Tron legacy.
  • Waifu Diffusion – Waifu Diffusion was trained on high quality anime images.
  • Wavyfusion – was trained on a diverse dataset ranging from photographs to paintings. It’s goal is to make a general purpose model for illustrated images.

Here’s a side-by-side of some images,so you can get a clearer picture of what each custom model does.

supermachine custom models
stable diffusion custom models

As much as I enjoy using regular Stable Diffusion, these custom models make creating stylish images A LOT easier. And you don’t have to use extremely long prompts to get impressive results – as the custom models do a lot of the heavy lifting.

If you have Supermachine already, be sure to try out the custom models for yourself. I’m sure you’ll enjoy them as much as I do.

If you don’t have Supermachine yet, you’ll find it here.