- ChatGPT Images launches a new, faster, and more accurate model, better able to follow user instructions.
- The update focuses on advanced editing, preset styles, and creative transformations from real photos.
- The system improves text within images, consistency in edits, and reduces the "yellowish" appearance of previous versions.
- The new version is integrated into ChatGPT and is available globally for both free and paid accounts.
Not too long ago, half the internet started playing at being an animated film director in the style of Studio Ghibli with the help of ChatGPT. Social media was flooded with dreamlike scenes, endearing characters, and backgrounds with that unmistakable Japanese touch. Behind this craze was OpenAI's image generation model integrated into ChatGPT, which turned anyone into a pseudo-illustrator without needing to know how to draw.
However, the world of generative artificial intelligence moves so fast that that trend was short-lived. Soon after, the conversation shifted to other offerings like Nano Banana Pro and Google's new models, capable of producing images so realistic they were difficult to distinguish from photographs, or tools like MidJourney that convert text into images. Faced with this landscape, OpenAI hasn't remained idle: it has unveiled a new version of ChatGPT Images designed to compete, offering greater speed, improved instruction comprehension, and a significant leap forward in advanced editing capabilities.
What is the new version of ChatGPT Images and what makes it different?
OpenAI has officially announced the next generation of ChatGPT Images, coinciding with the release of GPT-5.2, its latest language model. Both moves are part of the same strategy: to respond directly to Google's offensive with models like Gemini 3 Pro and demonstrate that they remain competitive at the highest level in text and image processing.
In this new phase, ChatGPT Images runs on the GPT Image 1.5 model , specifically designed to generate and edit images with greater control and speed than previous versions. The company itself claims that the system is now capable of producing results up to four times faster , a difference that is especially noticeable on mobile devices, where it was previously common for the process to be interrupted if the user switched applications mid-generation.
The comparison with the previous model is clear: where before there were long waiting times and a certain feeling of sluggishness, now the experience is more like a real-time digital studio . In the tests that have been shared, the images appear much more smoothly, with fewer freezes and without that impression that the app has frozen while the server calculates.
In addition to speed, OpenAI considers accuracy and adherence to user instructions to be the major qualitative leap. The goal is for ChatGPT Images to move beyond being a capricious generator that interprets prompts in its own way and become a more reliable tool for creative and professional workflows; that's why knowing how to write prompts for graphic design is crucial.
Advanced editing: change only what you want without ruining the image
One of the traditional Achilles' heels of image generators has been the localized editing of specific elements . If you wanted to retouch a jacket, change the background, or adjust the lighting, the model would often "reimagine" the entire scene, and you'd end up with a result completely different from what you intended.
The new version of ChatGPT Images focuses precisely on solving this problem. OpenAI states that the system is now able to modify only the parts specified by the user , leaving the rest of the image intact. This includes everything from specific objects (a car, a traffic sign, an article of clothing) to global aspects such as lighting, framing, and overall composition, without losing the overall coherence of the image.
The company explains that the model has been trained to:
- Modify only the requested elementsrespecting the original context as much as possible.
- Keep the rest of the image unchanged.preventing unwanted new details from appearing.
- Maintain consistency between the original image, the generated outputs, and the successive edits you apply.
This is especially noticeable in iterative workflows, where you make small adjustments over several rounds . In other models, after a few edits, the image would start to distort or lose quality: faces would change, backgrounds would become unrecognizable, details would disappear. With ChatGPT Images, the idea is that you can retouch, revert changes, fine-tune details, and combine versions without the final result degrading.
A good example is the type of prompt OpenAI has demonstrated: starting with the same urban scene, for example, a Los Angeles landscape with a person on a scooter, and requesting very specific changes: “make their shirt red, their cap yellow, the speed limit sign 15, and turn the truck into a fire truck.” Instead of regenerating everything from scratch, the model adjusts only those details, maintaining the original framing, atmosphere, and photographic style.
A “pocket-sized creative studio” for creating and reimagining scenes
OpenAI describes the new version as a kind of portable creative studio integrated directly into ChatGPT. The idea is that, whether you're starting from scratch or working from a real photo, you can guide the outcome much more precisely, almost like you would with a design team, but within a conversation.
The model excels at tasks such as adding or removing elements from a scene. For example, you can request the removal of an obstructing object, insert an additional character, add a parked car to the background, or remove people from a landscape without leaving any obvious traces of editing. The system reconstructs the background and adjusts shadows, reflections, and lighting so that the manipulation is seamless.
The ability to combine and merge images has also been enhanced . This allows, for example, taking a photo of a person and another of a completely different environment and generating a believable scene that blends both elements. Or combining subjects from several separate photographs into a single, coherent composition, something very useful for creating seemingly impossible but plausible montages.
Another important feature is the option to rearrange entire compositions . Through a prompt, you can request that certain elements of the frame be moved, that the camera zoom in or out, that the shot become a close-up or a panorama, or that the scene be reinterpreted at a different time of day. All of this is done while trying to preserve the visual characteristics that made the original image recognizable.
Finally, the transposition of styles and structures has been significantly refined . This means you can apply the composition, color palette, or lighting of a reference photo to another image, maintaining the structure of the original scene but with a different atmosphere. It's the kind of operation that allows you to go from a realistic scene to a cinematic aesthetic from the 2000s, or to a late 90s documentary street photograph shot on 35mm film with pronounced grain.
Creative transformations with your own photos
One area where the new ChatGPT Images particularly excels is in creative transformations using real images . Simply upload a photo of yourself (or someone else with your permission) and accompany it with a relatively simple prompt to obtain results in a matter of seconds that, in many cases, are surprisingly believable.
This idea isn't entirely new: models like Nano Banana had already placed great emphasis on this capability, allowing real portraits to be transformed into advertising campaigns, posters, or highly stylized scenes. What OpenAI is doing now is bringing that approach directly to ChatGPT, with a system that better understands the user's intent and more faithfully preserves the features of the original subject.
Among the most striking examples are prompts like: “Create an image of this man, but in Times Square in the winter of 2025.” The model adapts the clothing, surroundings, nighttime lighting, and urban elements to fit the aesthetic of New York at that time of year, while maintaining the gesture, general posture, and facial features of the original subject, and even allows the image to be converted into formats such as WhatsApp stickers.
Another common use is for visual "travel": you can request something like, "Place this person in full view in a Japanese city on a rainy night, with neon lights, reflections on the ground, and a cyberpunk aesthetic ." The system generates a scene with neon lights, puddles on the asphalt, and typical buildings, integrating the person into that world as if they were actually there, without them looking like a simple cutout pasted onto a background.
For those who prefer something more fantastical, transformations like "Turn this man into a Japanese samurai with traditional armor and katanas, in a realistic historical setting without modern elements" are possible . Here, ChatGPT Images reinterprets the clothing, posture, and context, but maintains a certain facial continuity so you still recognize the subject. The result might resemble a blend of historical illustration and recreated photography.
There's also room for nostalgic touches. One example OpenAI has shown is the request: "Use this man's image to create a full-body European advertisement from the 90s ." The model adjusts the color palette, framing, pose, and background to approximate the style of 90s advertising: period clothing, distinctive typography integrated into the scene, and that slightly dated feel that anyone who has seen ads from that era instantly recognizes.
Improved instruction tracking and more legible text in images
Another important improvement in the new ChatGPT Images is its ability to follow instructions precisely . Previously, it was common to ask for something very specific and receive a result that was similar but flawed in key details: incorrect colors, missing objects, or changes that hadn't been requested.
According to OpenAI, some of that problem is now a thing of the past. If you provide detailed instructions, the model should be able to follow them much more faithfully . This applies both to generating images from scratch and to editing an uploaded image: specifying the number of people, types of shots, specific colors, photographic or illustrative style, historical period, clothing, weather, etc., now has a more tangible impact on the final result.
The company also highlights specific improvements in text rendering within images . Historically, writing legible words on posters, signs, advertisements, or covers was one of the most challenging tasks for generative models, which often returned distorted characters, language mixes, or nonsensical words.
With this new version, ChatGPT Images is more competent at generating consistent typography and correct messages , which is key if you want to use the tool to create posters, promotional pieces, social media banners, or marketing materials where the text is as important as the background image.
Furthermore, work has been done on cases that are often particularly challenging, such as scenes with multiple small faces . In previous versions, when many people appeared in the background or in large groups, it was easy for faces to become distorted, repeated, or have unrealistic features. The new model improves the handling of these details, offering more natural results even in wide shots or crowd scenes.
Goodbye (almost) definitive to the yellowish tone of the early versions
For quite some time, one of the easiest ways to tell if an image had been generated with ChatGPT was its characteristic warm and creamy aesthetic . Many of these creations shared a palette dominated by yellows and oranges, soft lighting, and a "cinematic" feel that ended up becoming a kind of trademark, for better or for worse.
With the new version, OpenAI seems to have wanted to break with that unintentional visual signature. In the comparisons the company has published, the new model departs significantly from those default yellowish tones , unless the user explicitly requests them in the prompt. The images now tend toward a more neutral color representation or one more faithful to the user's input, without imposing such a pronounced style by default.
This has clear advantages for professional use: if you want a clean advertisement, a product mockup, or a photorealistic scene, you don't want the model applying a semi-permanent filter with that warm look that makes all the images seem identical. The new approach allows for much greater variation in the visual appearance without having to contend with an overly dominant base style.
In practice, this means it's now easier to achieve a wide range of results: from cool, desaturated images to vibrant, colorful aesthetics, including vintage styles, pastel tones, documentary photography, and hyper-bright advertising. The model is more flexible in adapting to the context and tone you specify in your instructions.
Predefined styles and suggestions: create without racking your brain with the prompt
One of the most welcome new features for those who don't want to struggle with complex prompts is the arrival of predefined styles and concepts within ChatGPT Images. Instead of writing exhaustive descriptions with technical parameters, you can directly choose a ready-made visual approach and let the system fill in the details.
These preset styles are designed to save time for both casual users and professionals who need consistency across different pieces. For example, you can select a specific aesthetic for a campaign and reuse it in multiple images without having to remember the exact prompt each time.
Additionally, within the images section of the ChatGPT app, you'll find a series of ready-to-use suggestions . You can start the process by choosing a style, uploading a photo, or even taking a selfie directly from the app, and let the AI suggest creative transformations without you having to think too much about the initial text.
This significantly lowers the barrier to entry for those unfamiliar with "the language of prompts." Instead of obsessing over describing everything down to the last detail, you can start with a reasonable base and then, as you see the results, request more specific adjustments.
In practice, ChatGPT functions here as an interactive creative assistant : it can suggest variations, propose style changes, framing adjustments, or new ideas related to the image you already have, so that the process ends up resembling less a simple one-off generation and more a collaborative art direction effort.
Where to use ChatGPT Images and how it integrates into the app
The new version of ChatGPT Images is being rolled out globally to all users , including those with free accounts. OpenAI has not announced any specific regional restrictions, so the idea is that anyone with access to ChatGPT can take advantage of the new image generation and editing capabilities.
If you use the mobile app, you've most likely already received a notification upon opening it, inviting you to start creating images . Additionally, the interface includes a new "Images" section in the sidebar, where you can view all your creations organized in one place.
This section serves not only as a personal gallery but also as a control center for managing your visual projects . From here, you can revisit old images to edit them again, duplicate results, compare versions, or start from a previous creation to experiment with entirely different styles without losing the original.
What's truly interesting is that all this functionality is integrated within the ChatGPT experience itself. You don't need to open external tools or export images to other editing services to make specific changes: you can request adjustments via text directly in the conversation , review the result, iterate again, and so on.
Ultimately, the new ChatGPT Images transforms the app into more than just a simple chatbot: it becomes a complete multimodal creation platform where text, image, and editing blend seamlessly, with a clear focus on covering both playful uses and more serious workflows, from quick advertising campaigns to product prototypes or visual sketches for presentations.
With this update, OpenAI aims to strengthen its position in the race for advanced visual generation, responding to the advances of Google and other players with a package that combines greater speed, improved accuracy, increased creative control, and seamless integration within ChatGPT. For the average user, the result is a tool for experimenting, testing ideas, and producing increasingly polished images without leaving the same chat where they already converse, schedule, or write on a daily basis.