Abstract:
OpenAI launches ChatGPT Images 2.5 and adds the Sketch function, which allows users to draw sketches in the chat box first, and then let the system generate more detailed images based on the sketches. OpenAI announced an update to ChatGPT's image generation tool on Tuesday. The core change is the addition of a new method called Sketch, which allows users to draw graffiti directly inside ChatGPT and then describe what kind of picture they want it to turn the sketch into.
When using it, just enter @Sketch in the chat box, and a drawing window will pop up. The user can draw the outline first, and then let the system continue to generate the finished product.
The report shows that The Verge reporter Jay Peters found after actual testing that the cat graffiti he drew casually with the mouse was successfully transformed into a realistic-style photo; he then directly marked the modification comments on the picture, changing the cat's eyes from blue to green, and then further adjusted them back to blue. This means that ChatGPT can now not only generate images based on sketches, but also support direct annotation modifications to parts of the image.
In addition to Sketch, OpenAI also introduced improvements to ChatGPT Images 2.5. According to officials, the images generated by the new version have "more natural light" and "richer textures", and are more capable of following multiple rounds of editing instructions. At the same time, the image generation delay can be reduced by up to 50% compared to Images 2.0. Currently, ChatGPT Images 2.5 is open to ChatGPT, ChatGPT Work and Codex users, supporting desktop, mobile and web pages.
Comments