OpenAI rolls out image generation powered by GPT-4o to ChatGPT
1 min read
Summary
ChatGPT is set to get image generation directly in the platform, using GPT-4, according to sources close to OpenAI.
The tool is set to be available across all versions of the platform, including the free version, but the number of daily image generations will be the same as with DALL-E.
The potential demise of DALL-E as a separate platform has been hinted at, with users likely to have access to the functionality via a custom GPT in the future.
The new image generator will use an autoregressive approach and will not have the same issues with text rendering and binding seen in other AI image generators.
Boundaries to prevent misuse such as Deepfakes and CSAM will be in place, but the new tool will not feature visual watermarks to indicate AI generation.
Research lead Gabriel Goh commented that “this model is a step change above previous models”.