OpenAI has announced the upgrade of its advanced image creation tool on the GPT-4o model, stating that it not only produces more beautiful images than before but also allows for more precise customization to meet specific requirements. Unlike its predecessor DALL·E, which focused on creating fixed images, the GPT-4o model operates incrementally, enabling the detailed specification and editing of images. Examples presented by OpenAI include pinpointing text within images, modifying images with both text and people, sequencing details up to 10-20 items in one prompt, learning from uploaded images, and matching text to create infographics, among others.
OpenAI indicates that the data used to train this image creation tool is sourced from publicly available information, including data from partners such as Shutterstock. The new image creation tool on the GPT-4o model is now available for Plus, Pro, Team, and free customers through ChatGPT, with limited daily usage similar to DALL·E. Enterprise and Edu customers will have access at a later date, with additional accessibility through Sora. For those still requiring the previous DALL·E tool, they can access it through Custom DALL·E GPT, while developers will have API access in the coming weeks.
TLDR: OpenAI upgrades its image creation tool on the GPT-4o model, offering enhanced customization capabilities and incremental processing for more precise image editing and creation. Publicly available data and partner sources are utilized for training, with accessibility available to different customer tiers through ChatGPT and Sora, along with API access for developers in the near future.
Leave a Comment