OpenAI Unveils DALL-E 2 With Advanced Image Editing Capabilities
OpenAI releases DALL-E 2, an upgraded text-to-image generator that boasts higher resolution and new editing tools. The system allows users to modify existing photos and create image variations based on text prompts.
OpenAI releases DALL-E 2, an upgraded version of its text-to-image generation model that delivers higher resolution and lower latency than the original. The new system builds on the CLIP computer vision framework to produce highly detailed and creative pictures based entirely on written descriptions. While the AI tool is not open-sourced, researchers currently have the opportunity to sign up and test its capabilities.
The updated model introduces a powerful "inpainting" feature that enables users to edit specific parts of an existing photograph using text commands. This means a person can highlight an area of a picture, such as a blank wall, and instruct the AI to add or replace objects seamlessly. The system even accounts for complex visual details like shadows when removing or adding elements to ensure a realistic final product.
DALL-E 2 also adds a variations feature, allowing users to upload a photo and generate alternative versions of that same image. Additionally, the AI can blend two separate pictures together to create a completely new image that incorporates elements from both sources. These advanced tools demonstrate significant progress in AI-driven creativity and image manipulation.