OpenAI has released a new flagship image generation model in ChatGPT that is available to all users, including those without a subscription. The model works up to four times faster and better understands user prompts when creating and editing images.
This was announced by OpenAI.
The new version of ChatGPT Images is rolling out today for all ChatGPT users and is also available via the API under the name GPT Image 1.5. The model allows users to create images from scratch or edit uploaded photos, precisely changing only the requested elements while preserving lighting, composition and people’s appearance.
OpenAI says the model has significantly improved its ability to follow instructions, render text within images, perform complex edits — such as adding, removing, blending and transforming elements — and produce more natural-looking results, including scenes with many faces and fine details. ChatGPT has also introduced a dedicated Images section with preset styles, filters and templates to help users quickly create visual content.
The new model differs from Sora, another OpenAI development, in both purpose and accessibility. ChatGPT Images is primarily designed for creating and editing static images directly in the chat and is widely available, including to free users with certain limitations. Sora, by contrast, is a more powerful multimodal model designed for complex tasks involving multiple types of content, including video, and is not yet as broadly accessible as image generation in ChatGPT.
OpenAI noted that while the new release represents meaningful progress compared to the previous version, the results are still not perfect, and further improvements are planned in future updates.