Openai brings AI image generation directly to its flagship chatbot. With ChatGpt’s 4O model, you can now create AI images locally in a normal ChatGpt window – without using DALL-E. This new feature is now available for free, with paid users available, with plans to bring it to next week’s business and education programs. You can also try image generation in your company’s AI video generator Sora.
So far, the businesses that OpenAi enters image generation have been focused on DALL-E, a separate program you can use via CHATGPT. You can use dall-e to create everything from sci-fi space opera to scenes shot in stock photography styles. DALL-E is our top pick for the best AI Image Generator, partly because its unique conversational style makes creating and editing images easy. Fortunately, this “chat” ability is coming soon. DALL-E is still available.
Chatgpt is available for free, with a paid plan starting at $20 per month. The current plan’s limitations will apply to image generation – if you are using a free plan, you can use the 4O model for limitations for messaging, file uploads, and data analysis. The same is true for Sora users. Chatgpt Plus users will get one image for each prompt.
Image generation in Chatgpt 4O will focus more on creating work-related images such as infographics and charts. Openai said this is improved text rendering to achieve this – which is very necessary when AI always hallucinates and messes up words in the image. You can also upload your own images and edit them using AI.
In the example image of OpenAI, the text in this image is very clear.
Chatgpt’s capabilities have some serious limitations. Most importantly, it says you may not be able to edit specific areas of an image accurately—the AI model can hallucinate things like 11 fingers, which is an essential task. If you upload your own image and edit it as the subject’s face, these changes may be lost from editing to editing. You may also see cropping and struggles related to data visualization and multilingual text. The company said in a blog post that it is working to improve these things and hopes to introduce fixes as early as next week.
Another example of Chatgpt’s image generation technique, with a pun with a snail-themed one.
Like dall-e, images made in chatgpt do not have any visible watermarks, indicating that they are generated by ai. Openai said its images will have C2PA metadata, an industry standard that lets people know that the images behind the scenes are made by AI. On the security side, OpenAI said it will follow the same content guidelines as the rest of the 4O model. It said its restrictions on nudity and graphic violence were “intensified.”
For more information, check out our complete review of CHATGPT and guide to writing the best AI image tips.
