OpenAI has launched ChatGPT Images 2.5, an update to the version 2.0 released in April. The new tool generates images up to 50% faster and features more natural lighting and richer textures. A key addition is the Sketch tool, which allows users to create rough layouts on touchscreens to serve as visual references for final image generation.
The update improves the preservation of subject details in reference photos, such as facial features, across multiple edits. The model now maintains stability in complex scenes with multiple objects or busy backgrounds, preventing the quality degradation that occurred in version 2.0. Previous requests in long conversations are now adhered to during successive edits without loss of quality.
The tool now includes templates for common formats like product photos and flyers, as well as support for inline comments for targeted edits. Users can also share prompts alongside generated images. OpenAI introduced two model variants: GPT-Image-2.5 Flare, the default high-speed model for brand assets and websites, and GPT-Image-2.5 Sunburst, which prioritizes precision and fidelity for professional marketing materials.
GPT-Image-2.5 Flare is reported to be two to four times faster than the GPT-Image-2 model and offers better support for transparent backgrounds. To ensure security, the system uses C2PA metadata and invisible watermarking to identify AI-generated content and prevent deepfakes. The new models are available in ChatGPT, ChatGPT Work, and Codex across desktop, mobile, and web platforms, with API access for developers.
