5 ms·

This is incorrect. Stable Diffusion 1.x was trained on "laion-improved-aesthetics" (a subset of laion2B-en).

• minimaxir 4y ago
  Double checked and both the initial comment and the correction are incorrect: the original v1.1 was trained on LAION-2B, then subsequent versions were finetuned on the aestethics subset. Either way, the main point is the same: more training data gives better results. https://github.com/CompVis/stable-diffusion#weights https://github.com/CompVis/stable-diffusion#weights
  • in3d 4y ago
    1.1 wasn’t public. Public releases were trained as I said.
    • cma 4y ago
      1.1 is available here: https://huggingface.co/CompVis/stable-diffusion-v-1-1-original https://huggingface.co/CompVis/stable-diffusion-v-1-1-origin...