OpenAI's newly released image generation model, ChatGPT Image 2.0, is shaking up the market with an explosive response immediately after its launch. As the model earns recognition for precision high enough for actual business use — beyond mere "picture generation" — user inflows are rapidly increasing.
"Now It's Really Usable"… User Surge Right After Launch
OpenAI said that as of the 30th of last month, daily active users (DAU) of ChatGPT Image 2.0 jumped more than 60% from the previous week in the first week after launch. New user inflows also surged more than 130%.
The model spread rapidly after being unveiled on the 21st (local time). In particular, as its performance rose to a level usable for "practical work" such as design drafts and content production, it drew reactions different from those of existing image generation AI.
In the industry, Google's image generation AI has led the way with realistic outputs and presentation-level polish, while OpenAI has been evaluated as relatively less accurate. However, with this update, there are forecasts that the competitive landscape could shift again.
Accurate Down to Text and Layout… "Penetrating Even the Design Domain"
The core of ChatGPT Image 2.0 is "accuracy."
It reflects users' detailed instructions more precisely and naturally implements the positional relationships, composition, and lighting of objects within an image. In particular, performance has been significantly improved in high-difficulty areas such as small text, icons, UI elements, and complex layouts.
The biggest change is text-handling capability.
The "broken text" problem frequently pointed out in existing image generation AI has been greatly reduced, and natural, readable text implementation is now possible even in non-Latin scripts such as Korean, Japanese, and Chinese.
In addition, up to 10 images can be generated at once, and screen ratios from 3:1 to 1:3 are supported, making the model usable for a variety of tasks including posters, diagrams, comics, and UI drafts.
"This model aims not for simple image generation, but for a level usable in actual production environments," OpenAI said.
"Shocked After Using It for an Hour"… Social Media Reactions Explode
Actual user reactions are also pouring in rapidly. On social networking services (SNS) both at home and abroad, evaluations that it is "finally at a level usable for real work" have continued.
Some users said, "This is really a big deal. OpenAI is back in the competition," showing reactions that the Google-centered market structure could change.
One user said, "I just ran Image 2.0 prompts for an hour, and the quality improvement is truly enormous. Complex scenes, accurate lighting, and consistent details — it really feels usable now."
Another user said, "For the first time, the image output quality is so good that it feels like I could actually use it for real work, not just for fun," mentioning its potential for practical use.
Korean User Characteristics… "Photo Conversion and Concept Utilization"
According to OpenAI's analysis of Korean user data over the most recent 24 hours, the most common use in the country was "converting existing photos."
Representative prompts included "Turn this photo into a high-end fashion studio pictorial," "Change it to an animation style," and "Make it into a cute mini-me character."
In other words, the model showed a characteristic of being used mainly for "reprocessing existing images" and "concept conversion" rather than simple generation.
"Asian users, including Koreans, are using it in very creative and practical ways," said Gabriel Goh, lead researcher at OpenAI. "The most interesting ideas often come from users."
