Openai integrates a new image generator in Chatgpt, which works on the basis of the 4O model. This also replaces the previous Dall-E model within Chatgpt-but which should continue to be available as a separate app.
This means that the omnimodal model 4O can now also see itself as a native image generator, since only this model is responsible for the generation of images and text.
Basically, Openai promises in theassociated presentationwho also asStream via YouTubewas held, a clear progress in three aspects:
- Precise text integration: The new image generator can insert texts much more precisely in images.
- Transparent background: Users can create images with transparent backgrounds.
- Improved context sensitivity: The model understands instructions better and can generate and modify images based on detailed descriptions or uploaded files.
As an example of the new possibilities, Openai mentions the photoUpload of a real living room; Based on this, visualized suggestions for the interior should follow. In addition, complex concepts or diagrams should look clearer and simply decent.
- Metadata should be used for the transparency in the context of the AI generated origin of such an image.
- The open standard C2PA (Coalition for Content Provenance and Authenticity) used, which is also used by other companies such as Adobe or Google to verify digital content.
Chatgpt-4o: More freedom in image generation
Openai-CEO Sam Altman also emphasized on theX-platformthat the new image generator should have less strict restrictions on the creation of content.
Users should receive "more freedom" to implement their creative ideas. At the same time, Altman said that Openaai would take social feedback into account to ensure that this freedom is used responsibly.
The new function is gradually introduced and is initially available to users with plus, pro, pro, team and free accounts. Enterprise and EDU customers should get access later.
