ChatGPT’s New Image Generator: Impressive, But Not Without Flaws
- Patrick Law
- Mar 27
- 2 min read
OpenAI has taken a bold step forward in the AI image space by integrating native image generation directly into ChatGPT, powered by its new GPT-4o model. This upgrade replaces DALL·E and introduces a more seamless, multimodal user experience — capable of producing images, text, and audio all in one chat.
But while early demos and user reactions have been overwhelmingly positive, there are still important limitations to keep in mind

What’s New
The new image feature enables users to create highly realistic or stylized images from simple prompts — no third-party tool needed. Within hours of its release, social media was flooded with Studio Ghibli–style portraits, surrealist compositions, and cinematic photo edits.
Testers from Decrypt put the model through rigorous evaluation and found it could accurately render 23 of 25 elements in a complex scene — outperforming leading image generators like Reve and Flux in prompt comprehension and spatial awareness.
Users can also edit images using natural language, including turning photos into themed posters or modifying specific parts of an image with precision — a major usability advantage over traditional inpainting tools.
What to Watch Out For
Despite the progress, the feature isn’t without issues:
Moderation gaps still exist. According to the New York Post, testers were able to bypass safety filters by simply rewording prompts — raising concerns about content control.
As The Verge reports, images contain C2PA metadata but no visible watermark, making them harder to identify as AI-generated. This increases the risk of misuse or misinformation.
The rollout has also hit capacity issues. CEO Sam Altman confirmed on X that demand was higher than expected, resulting in a delay for free-tier users. As of now, only Plus, Pro, and Teams subscribers have access.
Verdict
There’s no doubt that ChatGPT’s new image tool is powerful. It rivals — and in some areas, surpasses — dedicated platforms in both ease of use and accuracy. For many users, it now offers the most complete and accessible creative toolset for just $20/month.
However, moderation flaws, lack of visible transparency, and limited rollout show there’s still work to be done before this becomes the new standard.
Check out the video.
Sources:
https://techcrunch.com/2025/03/26/chatgpts-new-ai-image-feature-is-delayed-for-free-users/
https://www.nbcnews.com/tech/tech-news/studio-ghibli-chatgpt-inspired-memes-portraits-rcna198273
https://decrypt.co/311761/review-openais-new-image-generator https://nypost.com/2025/03/26/business/chatgpt-wont-generate-images-of-sexy-women-but-will-make-images-of-sexy-men-due-to-apparent-bug/
Comments