OpenAI推出ChatGPT Images 2.5，延迟减半并解锁草图微调

OpenAI Launches ChatGPT Images 2.5 with Delay Reduction by Half and Sketch Fine-Tuning Unlocked AIbase基地 Published inAI News · 4 min read · Sep 9, 2026 21 OpenAI has officially launched its most advanced AI image generation model to date - ChatGPT Images2.5. Compared to Images2 released in April of this year, the new version significantly improves image detail sharpness, natural lighting, and texture richness while reducing image generation latency by up to 50%, making the entire creative process more efficient and smooth. In terms of functional innovations, Images2.5 brings several major upgrades tailored for practical scenarios. The most notable is the brand-new Sketch feature. Users can call it by entering the "@Sketch" command and directly draw a sketch and add text descriptions in the conversation, which the model will then convert into a complete detailed image. In addition, the system includes templates for common formats such as posters and products, greatly lowering the barrier to creation. Users can also attach the original prompt words with one click when sharing images. For professional design and detailed modification needs, the new version introduces a comment annotation feature within images. Users can directly add annotations at specific locations on the image and issue instructions such as "delete this area" or "change this to blue." During this process, the model only makes precise modifications to the specified area, keeping the rest unchanged. It effectively maintains image consistency and quality across multiple rounds of editing, avoiding issues like detail degradation that often occur with traditional multi-round modifications. Regarding API calls, OpenAI has introduced two models with different focuses for developers to choose from. GPT-Image-2.5Flare is aimed at most general API image generation applications, suitable for scenarios such as creator content, social media, product experience, visual search, rapid prototyping, and high-volume generation; while GPT-Image-2.5Sunburst is specifically designed for complex workflows requiring higher editing control, making it more suitable for finished marketing materials and refined product images, although the generation time is relatively longer. The API supports various resolutions, with the longest side not exceeding 3840 pixels and total pixel count between 655,360 and 8,294,400. In terms of security and compliance, Images2.5 continues the existing security mechanisms, continuously checks prompts and generated images to prevent harmful outputs. At the same time, all output content is embedded with C2PA metadata and invisible watermarks, making it easier for external parties to identify images generated by OpenAI tools, thus effectively combating unauthorized reuse and deepfake behaviors. ChatGPTImages2.5AI Image Generation ModelSketch FeatureOpenAI This article is from AIbase Daily Scan to view Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications. —— Created by the AIbase Daily Team © Copyright AIbase Base 2024, Click to View Source -