Skip to content

Step1X-Edit

A unified image editing framework driven by instructions

Visit official site

What Step1X-Edit is

The framework is described as a practical general-purpose image editing framework that uses the image understanding of MLLMs to parse editing instructions, generate editing tokens and decode them into images through a DiT network. It is presented as a unified image editing model that supports many user instructions and addresses real editing needs.

What you can do with it

  • Support a wide range of image editing instructions
  • Apply advanced machine learning techniques to improve editing accuracy
  • Use the GEdit-Bench benchmark for evaluation in real scenarios
  • Work with various image formats and open source code

Example uses

  • A designer quickly adjusts product images for promotion
  • A social media creator edits images through simple instructions
  • A general user makes simple adjustments and beautifies family photos

Getting started

  • Visit the official Step1X-Edit page
  • Download the model weights and inference code
  • Set editing instructions based on the provided technical report
  • Decode editing tokens with the DiT network

Pros & cons

✓ What we like

  • A unified model supports many kinds of user instructions
  • Open source code allows secondary development and customisation

! What to watch out for

  • Editing runs through instruction parsing, token generation and DiT decoding
  • Getting started requires downloading weights and inference code

FAQ

How are editing instructions handled?

MLLMs parse the instructions, editing tokens are generated, and a DiT network decodes them into images.

Is the source available?

The feature list states the code is open for secondary development and customisation.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review