Instant AI Art

Imagine, draw & generate: see your new artwork come to life in an instant

Comment 1

Inspiration

  • Easier access to creativity

What it does

  • Take an existing image and transform it with strokes and a text description

How we built it

  • Started with the Canva Apps SDK using the image_editing example, tested it thoroughly and integrated the stable diffusion model using a backend API
  • @Imagineers AI dev did the rest

Challenges we ran into

Challenges, Technical

  • sharing states between the object panel (iframe) and the image canvas
  • choosing the color of the stroke so it can better blend with the image
  • selecting the optimal the brush size
  • finding an on-demand GPU cloud to lower costs before the app goes live

UI challenges

  • Generation trigger (after clicking on canvas or primary button)
  • how to handle strokes if the user drags outside of overlay canvas

Accomplishments that we're proud of

  • Integrating Overlays API with image generation model
  • Seamless interface to edit and recreate images
  • Low cost model hosting during review process
  • Providing design/creative opportunity to broader community

What's next for Instant AI Art

  • Undo/Redo buttons (implemented before submission)
  • An interface to view all prior generations and choose any one of them as a template
  • Better output quality with more fine-grain control for enterprise use cases

Built With

  • canva-apps-sdk
  • diffusers
  • diffusion
  • fastapi
  • gpu-cloud
  • jwt-auth
  • model
  • overlays-api
  • postgresql
  • python
  • pytorch
  • react-typescript
  • selection-api
  • torch
Share this project:

Updates