ChatGPT Images 2.5 Adds Sketch Canvas and Transparent Backgrounds

SAN FRANCISCO — OpenAI has officially deployed ChatGPT Images 2.5, an update that brings native transparent background rendering and interactive drawing tools to digital asset creation. It also introduces targeted enterprise API tiers for developer integration.

The model is available through two Application Programming Interface (API) endpoints: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. Flare corresponds directly to the default model integrated into the end-user ChatGPT interface, while Sunburst is an expanded, high-capability alternative for complex visual instructions and advanced generation tasks.

OpenAI has kept developer pricing unchanged from the Images 2.0 generation released in April. API access remains set at up to $8.00 per million input tokens and $30.00 per million output tokens, allowing enterprise clients to scale automated image pipelines and visual content generation without increased per-token infrastructure expenses.

Processing latency has been cut in half compared to the April release. The 50 percent speed increase shortens queue times during high-volume usage without reducing output resolution or visual clarity.

The 2.5 release also upgrades prompt adherence, surface texture detail, and atmospheric lighting mechanics. Across multi-image prompts, the model has improved precision when extracting and preserving facial features or reference subjects from real-world photographic inputs.

Native support for transparent backgrounds is another addition aimed at commercial design workflows. Earlier visual AI iterations typically placed subject matter against solid or complex backgrounds, which required manual clipping paths or third-party background removal software. Designers can now export isolated graphical elements directly into layout applications, brand toolkits, and software interfaces.

OpenAI has also added Sketch to the ChatGPT interface. Typing `@sketch` into the chat prompt opens a dedicated canvas within the conversation window, where users can work with basic drawing and color controls through a mouse or touch interface.

The model processes user-drawn outlines or color maps as structural references, turning rough sketches into refined graphics, characters, or photographic renders in seconds. Users who do not have existing art assets or complex text prompts can instead begin with pre-designed layout templates.

Those starter frameworks include brand logos and vector-style graphics, architectural interior design concepts, marketing materials and promotional posters, and product design and merchandise mockups.

ChatGPT Images 2.5 is rolling out across OpenAI’s entire product lineup. Free-tier ChatGPT users and paid subscribers can access the model, which is also integrated across developer and professional workspace environments, including Codex and ChatGPT Work.

The update arrives as the generative AI sector moves toward multimodal, interactive canvas environments. Modern creation tools increasingly combine freehand drawing, layout templates, real-world reference imagery, and developer API endpoints into unified design systems.