OpenAI introduces ChatGPT Images 2.5 with faster, more precise editing

OpenAI has released ChatGPT Images 2.5, an updated image generation and editing system that the company says improves speed, subject fidelity, and control over detailed revisions. The rollout includes new creative tools in ChatGPT, such as a drawing feature called Sketch, image comments, and templates for common formats including posters and product images.

According to OpenAI, Images 2.5 can reduce generation latency by up to 50 percent compared with Images 2.0. The company says the model also produces more natural lighting and textures, and does a better job retaining recognizable characteristics when users work from reference photographs.

The update is available in ChatGPT, ChatGPT Work, and Codex across desktop, mobile, and web. Developers can access two related models through the API: GPT-Image-2.5 Flare, intended as the default option for most applications, and GPT-Image-2.5 Sunburst, which targets more detailed and slower premium creative workflows.

More control over edits

A central focus of the release is targeted image editing. OpenAI says the new model is more likely to change only the requested part of an image while preserving the subject, layout, background, and visual treatment around it. That could be useful for marketing teams that need to revise product details, text, or backgrounds without rebuilding an entire asset.

The company also says Images 2.5 retains prior changes more reliably during multi-step editing sessions. In practice, this is meant to reduce a common problem in generative image tools: earlier edits disappearing or the overall image drifting after several rounds of instructions.

Ina Fried reports for Axios that early testing found stronger preservation of people and pets in edited images. Axios tested the system on a cat photo, a tattoo concept built around an existing tattoo, and a logo design. The report found that users could make revisions by combining written instructions with visual annotations, including circles drawn around areas that should change.

Sketches, templates, and shared prompts

Sketch lets users draw directly in ChatGPT and use the drawing as a visual reference. A rough layout, object outline, or doodle can be combined with a written prompt that specifies style, materials, colors, or other details. Users can activate the tool by typing “@Sketch,” OpenAI says.

Templates offer starting points for selected creative tasks. Instead of writing every instruction from scratch, users can select a format such as a poster or merchandise design and fill in the relevant information. ChatGPT also now supports placing comments directly on an image, which can help direct an edit to a specific location.

Another new sharing option lets users include the prompt when they share an image. Other people can reuse that prompt and adapt it with their own photos or details.

The improvements arrive as image generation tools face continued scrutiny over training data, artist rights, and the difficulty of identifying synthetic visuals. OpenAI says Images 2.5 continues to use prompt and image safety checks, C2PA metadata, and invisible watermarking. Axios notes that the output can still show signs of AI generation, although those signs are becoming harder to spot.

Sources

• Introducing ChatGPT Images 2.5 – OpenAI
• OpenAI launches ChatGPT Images 2.5, which it says cuts image generation latency by up to 50% vs. Images 2.0, and adds a Sketch feature for drawing in ChatGPT – Axios