cjwbw / portraitplus

Portraits with stable-diffusion

  • Public
  • 24K runs

Run time and cost

This model costs approximately $0.027 to run on Replicate, or 37 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia T4 GPU hardware. Predictions typically complete within 120 seconds. The predict time for this model varies significantly based on the inputs.

Readme

weights from: https://huggingface.co/wavymulder/portraitplus

Portrait+ Header CKPT DOWNLOAD LINK - this is a dreambooth model trained on a diverse set of close to medium range portraits of people.

Use portrait+ style in your prompt (I recommend at the start)

The goal was to create a model with a consistent portrait composition and consistent eyes. See the batch example below for the consistency of the model’s eyes. This model can do several styles, so you’ll want to guide it along depending on your goals. Note below in the document that prompting celebrities works a bit differently than prompting generic characters, since real people have a more photoreal presence in the base 1.5 model. Also note that fantasy concepts, like cyberpunk people or wizards, will require more rigid prompting for photoreal styles than something common like a person in a park.

Portrait+ works best at a 1:1 aspect ratio, though I’ve had success with tall aspect ratios as well.

Please see this document where I share the parameters (prompt, sampler, seed, etc.) used for all example images above.

Modification example

Batch example