GitHub Copilot can now generate and edit images from natural-language requests, including during a coding task. GitHub's October 28, 2026 changelog says the selected reasoning model calls a dedicated image-generation tool powered by GPT Image 2.5 Flare. The result is previewed in the conversation and saved to the project, keeping the visual asset within the same workflow.
The capability is available in Copilot chat in VS Code, the GitHub Copilot app and Copilot CLI. It is in public preview for Pro, Pro+, Business and Enterprise plans. That does not establish free-tier access or final general availability. Users evaluating it should check both their plan and organizational policy if their account is managed.
The chat model calls a separate image tool
The distinction between reasoning model and image tool explains the workflow: the user continues an existing conversation, and the model delegates image creation to the dedicated tool. A separate request is unnecessary simply because a task now needs a visual asset. Users can describe the image during their work, preview the result and find it in the project workspace.
To make the asset fit, specify its purpose and important characteristics, such as composition, colors or elements that must remain during editing. The description need not be lengthy, but it should make the outcome assessable. An image saved to a project still needs review: does it fit the design, preserve important details and answer the actual request?
Image-generation billing is separate from chat-model billing. It is based on image tokens that count toward the plan's AI credits. An open chat therefore does not mean a visual operation carries no additional cost. GitHub's update does not state a single fixed price for every image, and no such price should be attached without the relevant pricing details.
Organizational controls and image labeling
Organization and Enterprise owners can turn off image generation through Copilot model policies. That is the control described for this update, relevant to teams deciding which tools may run in their workspace. A capability's presence in the product does not mean an organization has enabled it for every user.
GitHub also offers a visible "Made with AI" watermark for generated and edited images. It is off by default and can be enabled voluntarily. An absent watermark therefore does not prove an image was not AI-generated. For developers, the change integrates creation, preview and saving into one sequence. Budget, usage policy and asset review remain part of deciding how to use it.