If you've just opened X and happen to be following OpenAI and Sam Altman, you might see a picture like this:
Yes, it's definitely a bit of an eyesore. Even the official OpenAI account couldn't help but comment (actually, it was just to get attention): sam.
The comments section was filled with complaints and jokes:
But regardless, it certainly generated buzz.
In fact, the reason Sam Ultraman posted such an eye-catching image was to promote OpenAI's newly launched ChatGPT Images . And this isn't the only image based on an Ultraman photo. OpenAI's official account even posted a separate image like this:
And he's not the only OpenAI executive to have personally contributed photos. OpenAI Chief Research Officer Mark Chen and OpenAI President Greg Brockman have both shared their images. For example, below is a screenshot of their edited conversation shared on their official blog:
New ChatGPT Images
The new version of ChatGPT Images is powered by OpenAI's new flagship image generation model. The company states, "Now, whether you're creating or editing a photo from scratch, you can get the image you envision. It preserves detail while allowing for precise editing, and image generation is up to 4 times faster."
Its core feature is: precise editing, preserving key details .
Now, when users request editing of uploaded images, the model can more reliably follow the user's intentions. According to OpenAI, this is "meticulous down to every little detail," such as changing only the parts requested by the user while maintaining consistency in elements such as lighting, composition, and the appearance of the subject across input, output, and subsequent editing.
This unlocks many potential use cases, such as more practical photo editing, more realistic clothing and hairstyle try-ons, or style filters and concept transformations that preserve the essence of the original image.
Editing function
OpenAI states that the model excels at various types of editing, including adding, deleting, combining, merging, and replacing.
Below is an example from generation to various editing processes.
Creative Transformation
The model also features transformation capabilities, allowing users to change and add elements (such as text and layout) to realize ideas while preserving important details. OpenAI states that these transformations work for both simple and complex concepts and can be easily experimented with preset styles and creative ideas in the new ChatGPT Images feature—without the need to manually write prompts.
For example, you can turn an everyday photo into a movie poster.
Or make a picture of Ultraman wearing a pearl earring.
Follow instructions
OpenAI states, "Compared to our initial version, this model follows instructions more reliably. This not only enables more precise editing but also handles more complex original compositions, preserving the relationships between elements as expected." A comparison example is provided below:
Text rendering
This model represents a significant step forward in text rendering, enabling it to handle denser and smaller text.
Other quality improvements
The model has also been improved in other dimensions, making the output more readily usable, such as rendering many small faces and improving the naturalness of the output.
A brand new creative space
In addition, OpenAI has launched a brand-new " Images " feature in ChatGPT, "aimed at making the image generation experience more enjoyable—not only inspiring, but also making creative exploration effortless."
Users can access it via the mobile app's sidebar and chatgpt.com, making image exploration and experimentation faster and simpler. It includes dozens of preset filters and prompts to quickly spark inspiration and is regularly updated to reflect emerging trends.
Availability
The all-new Images model is available to all ChatGPT users starting today, and is offered via API as GPT Image 1.5. Compared to GPT Image 1, GPT Image 1.5 reduces image input and output costs by 20%, allowing users to generate and iterate on more images without increasing their budget.
The new Images experience in ChatGPT is also available to most users today, with Business and Enterprise access to be rolled out later.
OpenAI stated, "We believe we are still at the beginning of the possibilities that image generation can achieve. Today's update is a significant step forward, with more features to come, from finer-grained editing to richer, more detailed cross-language output."
Conclusion
Finally, we also used the same prompt to conduct a simple test comparing ChatGPT Images with Google's Nano Banana Pro.
Results from ChatGPT Images
Results from Nano Banana Pro
It appears that, at least in this specific case, the results from ChatGPT Images are more in line with our instructions.
In summary, in this battle of AI-generated images, OpenAI's release of ChatGPT Images signifies a shift in the competitive landscape: from simple "model capabilities" to a comprehensive "product experience." OpenAI didn't even release any quantitative benchmark results this time!
Meanwhile, Sam Altman's self-deprecating humor, while full of Silicon Valley-style wit, sends a very clear signal: OpenAI is making every effort to lower the psychological barrier for users. By launching a separate "Images" entry, pre-set style filters, and significant price reductions, OpenAI is attempting to transform image generation from a professional skill requiring carefully tuned prompts into an everyday interaction as simple as tweeting.
Do you have a positive outlook on ChatGPT Images?
Reference Links
https://openai.com/index/new-chatgpt-images-is-here/
https://x.com/OpenAI/status/2000990989629161873
This article is from the WeChat public account "Machine Heart" (ID: almosthuman2014) , author: Panda, and published with authorization from 36Kr.
