Skip to content

HyperDreamBooth

Personalize a text-to-image model from a single face image

Visit official site

What HyperDreamBooth is

The model is described as a hypernetwork developed by Google Research for fast personalization of text-to-image models. The model generates a small set of personalized weights from a single face image and uses fast fine-tuning to produce face images across contexts and styles.

What you can do with it

  • Generate personalized weights from one portrait with a hypernetwork
  • Combine the weights into a diffusion model for fast fine-tuning
  • Produce face images across varied contexts and styles
  • Keep the model knowledge for diverse styles and semantic edits

Getting started

  • Prepare one clear face image of the target person
  • Visit the project web page
  • Upload the face image to the model
  • Choose the desired style and context

Model notes

  • Personalization is stated to finish in about 20 seconds
  • The material calls it 25 times faster than DreamBooth
  • The resulting model is stated to be 10000 times smaller than a regular DreamBooth model
  • Quality and style variety are stated to match DreamBooth

Pros & cons

✓ What we like

  • Only one reference face image is needed
  • Fine-tuning is described as rapid
  • The material states quality and style variety match DreamBooth

! What to watch out for

  • The documented input is a single clear face image of the target person
  • Personalization is described through a hypernetwork combined with a diffusion model
  • The documented flow requires visiting the project page and uploading an image

FAQ

How many images are needed?

The material states only one reference face image is used.

How fast is personalization?

The material states about 20 seconds, described as 25 times faster than DreamBooth.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review