ChatGPT Can Now Create Better Images: OpenAI Brings a New Generation of AI Images Directly to ChatGPT
OpenAI has introduced new image-generation capabilities powered by artificial intelligence directly in ChatGPT and Sora through its GPT-4o model. This innovation allows users to create and edit images directly in the chat interface, eliminating the need for external tools such as DALL-E.
What Does the New "Images in ChatGPT" Feature Bring?
The new "Images in ChatGPT" feature powered by the GPT-4o model brings several significant improvements:
Creating and editing images directly in the chat - Users can generate new images or edit existing ones using natural language or uploaded files directly in ChatGPT
Available to everyone - The feature is available to all users (Free, Plus, Team, and Pro) without strict limits for free users (restrictions may be adjusted according to demand)
Improved text rendering - Significant improvement in the accuracy of text rendering in images
More accurate colors and textures - Generating photorealistic images with precise lighting effects and textures
Why Is This Important?
OpenAI has long believed that image generation should be a primary capability of its language models. People have always used visual communication - from cave paintings to modern infographics - to convey information, persuade, and analyze, not merely for decoration.
While previous generative models were able to create impressive surrealistic scenes, they often failed when creating practical images that people use to share and create information. GPT-4o addresses these shortcomings:
Significantly better consistency - ChatGPT can now accurately place 15 to 20 elements in a single image, compared with just 5 to 8 elements in previous models
Excellent creation of structured visuals - The model excels at generating menus, diagrams, and infographics with readable text
Improved image editing - Enables modifications to existing images (including those containing people) while preserving complex scenes
How Did OpenAI Achieve This?
OpenAI used a group of human trainers who labeled training data for the model, enabling:
Generation of more accurately rendered and useful images
Better adherence to human instructions
Using the model's extensive knowledge to better understand context
Thanks to GPT-4o's multimodal capabilities, you can now create exactly the images you envision and communicate more effectively through visuals.
This improvement replaces OpenAI's previous DALL-E model and brings long-awaited advances in text rendering, design capabilities, and natural-language editing. It represents a new era of AI-generated visual content that is now more accessible and intuitive for everyone.
You can watch a demonstration video here.
