OpenAI Unveils DALL-E 3 Image Generator – to Be Integrated into ChatGPT
OpenAI has introduced a new version of its image generator, DALL-E 3, and announced plans to integrate it into ChatGPT. Developers state that DALL-E 3 can “significantly better understand requests,” analyze complex instructions, and generate “extremely detailed and accurate images” compared to DALL-E 2.
“Modern image generators tend to ignore words or descriptions, forcing users to engage in prompt engineering. DALL-E 3 is a step forward in our ability to create images that precisely match your request,” says OpenAI in the description of the new image generator.
The model also performs better with intricate details, such as human hands. Even with identical requests, DALL-E 3’s results are far superior to those of DALL-E 2, according to the developers. DALL-E 3 can accurately depict scenes with specific objects and consider how they relate to one another, as shown in the image below.
OpenAI plans to integrate DALL-E 3 into the ChatGPT Plus and Enterprise chatbot offerings in the near future. Combining the language capabilities of the chatbot with an image generator will allow for even more precise image creation and provide more options for fine-tuning a request if the initial result does not meet the user’s expectations.
“When requested, ChatGPT will automatically generate detailed prompts for DALL-E 3 based on which it will create the image. If the generated image is liked by the user but doesn’t quite match the request, you can make adjustments by adding a few clarifying words,” OpenAI explains.
Currently, DALL-E 3 is in the research preview stage and will be available to ChatGPT Plus and Enterprise subscribers in October via the API. It’s worth noting that the use of DALL-E 2 is paid, and a monthly subscription to ChatGPT Plus costs $20. Currently, the only major AI chatbot offering a free integrated image generator is Microsoft’s Bing Chat AI, which operates on the powerful GPT-4 language model from OpenAI.
“Similar to DALL-E 2, images generated with DALL-E 3 will belong to the user, and they won’t require permission to reproduce, sell, or distribute them,” clarifies OpenAI.
The company also highlighted that DALL-E 3 includes tools to restrict the creation of certain images. For instance, the image generator will reject requests involving public figures or “harmful biases” to reduce the risk of propaganda and misinformation spreading. Developers are also testing an origin classifier, a tool that can determine whether an image was generated by an AI model. Lastly, DALL-E will not generate images in the style of currently living and active artists.
- I'm Martin Harris, a tech writer with extensive experience, contributing to global publications. Trained in Computer Science, I merged my technical know-how with writing, becoming a technology journalist. I've covered diverse topics like AI and consumer electronics, contributing to top tech platforms. I participate in tech events for knowledge updating. Besides writing, I enjoy reading, photography, and aim to clarify technology's complexities to readers.
- Unusual04/11/2023Japanese Scientists Develop Durable, Self-Healing, Biodegradable Plastic
- Blockchain03/11/2023Telegram’s TON Becomes World’s Fastest Blockchain, Outpaces Visa and MasterCard
- Internet27/10/2023Google Search Now Provides Source Data for Images
- Business26/10/2023Honor Leads Chinese Smartphone Market as vivo Falters, Apple Reclaims Top Three