Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
ChatGPT’s image generation capabilities just got an upgrade. OpenAI posted on its account on the social network X an update declaring that users of GPT-4 powered ChatGPT can now click on images generated within the chatbot and use a paintbrush style tool to highlight sections of the image they want to adjust or add to.
In a video showing off the capabilities, a user generates images of a dog with a conical party hat atop it.
The user can then be seen clicking the image to expand it and enter a new editor view, where they select a paintbrush icon at the top, brush over portions of the dog’s head, and then type in a new prompt “add bows,” which preserves the original image but generates a new version with red bows appearing in the user-highlighted region.
Several web users noted they had observed the changes pushed live quietly earlier this week and last.
AI Scaling Hits Its Limits
Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:
- Turning energy into a strategic advantage
- Architecting efficient inference for real throughput gains
- Unlocking competitive ROI with sustainable AI systems
Secure your spot to stay ahead: https://bit.ly/4mwGngO
VentureBeat tried out the new feature on a personal account and it worked as described.
Powered by DALL-E 3 and another new update
OpenAI’s image generation feature within ChatGPT is powered by its underlying DALL-E 3 model which was unveiled by the company back in September 2023. The addition of DALL-E 3 directly within ChatGPT greatly expanded the latter model’s capabilities for users, and the new addition today again further extends its utility and fine-grained control.
VentureBeat users ChatGPT/DALL-E 3 and other AI art generators to create article header images and other artwork for our properties.
OpenAI also tweeted that it has added a new style inspiration feature in its DALL-E 3 GPT, allowing users to select between several pre-set styles such as woodcut, close-up, low angle, artificial lighting and hand-drawn.
The moves come even amid ongoing legal challenges to OpenAI’s models and growing pushback by some artists and web users against AI art generally, especially models like DALL-E 3 that were presumably trained on copyrighted imagery without express permission, consent or compensation.
