replicate / dreambooth

Train your own custom Stable Diffusion model using a small set of images

  • Public
  • 291.9K runs
  • GitHub
  • Paper
  • License
  1. a8ba568d

    Latest

    Release notes

    Updated: Dreambooth Trainer that trains on top of an existing custom CKPT.

    This version improves compatibility with many CKPT files that failed previously.

    You must provide a URL for ckpt_base.

    Users have reported success with checkpoints such as Analog Diffusion, OpenJourney-v2, Elden Ring Diffusion, and Deliberate.

  2. Release notes

    deprecated version: Dreambooth Trainer that allows you to start from custom CKPT file.

    This version fixes issues with “bare” ckpt files reporting an error about missing state_dict

    You must provide a url for ckpt_base, training starts from an existing fine-tuned checkpoint

    Users have reported success with checkpoints such as Analog Diffusion, OpenJourney-v2, Elden Ring Diffusion.

  3. Release notes

    Dreambooth Trainer based on Stable Diffusion 2.1-base (512x512 version)

    See blog post for how to use with our experimental Dreambooth API and pin to this trainer_version

  4. Release notes

    this is a deprecated version with ckpt_base support, please use the more recent version.

  5. Release notes

    Dreambooth built on Stable Diffusion 1.5

    This is the default version used when using the Dreambooth API if you don’t specify a trainer_version