Skip to content
Get packs

AR Cut & Paste

A research prototype that lets your phone cut objects from the real world and paste them into Photoshop.

GitHub Download source
Stars
14.6k
Forks
2k
Licence
MIT
Language
TypeScript
Last update
3.7 years ago
Type
Other

No commits in over a year — check for forks or alternatives before adopting.

What it is

AR Cut & Paste is a research prototype that uses a phone camera and machine learning to pick out an object from your surroundings and paste it into Photoshop. It runs as three parts: a mobile app, a local server that bridges the app and Photoshop, and an external service that detects the object and removes its background. The README describes it as a prototype, not a consumer tool, and points to a separate app, clipdrop.co, for easier use.

  • Photograph an object with the phone app and send its cutout to an open Photoshop document
  • Study how the local server finds a screen position from the camera using the linked screenpoint project
  • Swap in a different background removal service by changing the server configuration

Great for

  • Developers curious about combining AR, image segmentation and Photoshop automation
  • Researchers exploring salient object detection (BASNet) in a practical pipeline
  • People who already use Photoshop and can set up a local server

Not ideal for

  • Everyday users wanting a ready-made app: the README says it is a research prototype and points to clipdrop.co instead
  • Users without a CUDA GPU who cannot rely on the community-hosted endpoint, since the README gives no uptime guarantee
  • Anyone needing non-Photoshop editors: the README says only Photoshop is handled currently
Step-by-step guide: how to use AR Cut & PasteInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    Clone and build from source

    You'll need: A CUDA GPU (only if self-hosting the BASNet service) · Photoshop with Remote Connection enabled · Python (the local server uses pip and requirements.txt)

    Clone the repository

    git clone https://github.com/cyrildiagne/ar-cutpaste
  2. 2

    Run it

    • Enable Remote Connection in Photoshop under Preferences > Plug-ins and set a password
    • Match your Photoshop document settings to server/src/ps.py
    • Set up the salience detection service, either your own BASNet deployment or the community endpoint
    • Follow the /server and /app folder instructions to run the local server and deploy the mobile app
  3. 3

    Get the most out of it

    • Use a document with a visible background, since the README says a blank background may stop SIFT from matching correctly
    • Run BASNet on a separate port if it shares a computer with the local server
    • Pass the community endpoint with the --basnet_service_ip flag when launching the local server

Downloads

Source code (.zip)master branch · straight from GitHub

Strengths

  • Open source under the MIT license, so the code can be read and reused
  • Clearly split into three modules, which makes individual parts easy to inspect or replace
  • Documents its own limits honestly, including that it is a prototype

Trade-offs

  • Its last push was 2023-03-04 and the README describes it as a prototype, so it may not work with current software
  • Depends on an external segmentation service; the self-hosted option needs a CUDA GPU and the public one is community-provided
  • Only Photoshop is supported, and setup requires configuring Photoshop settings to match the server code

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.