Ddos 
Developer @btibor91 on the social platform X (Twitter) has discovered that OpenAI appears to be preparing to introduce watermarking to images generated by its GPT-4o model on the Android version of ChatGPT. The watermark, labeled “ImageGen,” is intended to signify that the image is synthetically generated rather than authentic. However, it remains unclear whether this will be a visible (overt) watermark or a concealed (invisible) one.
An increasing number of users have begun leveraging ChatGPT to create images in the distinctive style of Studio Ghibli animation. Indeed, the viral surge of GPT-4o’s popularity owes much to its capacity to produce Ghibli-style imagery, which has rapidly proliferated across social media. OpenAI has revealed that users have already generated over 700 million images using the model.
Following this wave of enthusiasm, OpenAI has started to implement copyright safeguards. Currently, if users explicitly request an image to be redrawn in the Ghibli style—or to generate new images in that aesthetic—the system will return a copyright error. However, users can still upload an existing Ghibli image and ask GPT-4o to recreate their own image in the same visual style, effectively circumventing the restriction.
Initially, the image generation feature was exclusive to ChatGPT Plus subscribers. Now, however, it is accessible to free-tier users as well—albeit with limitations. Free users may generate up to three images per day, and failed attempts still count toward the quota. Once the limit is reached, users must wait 24 hours for the allowance to reset.
According to code inspection, the watermark is likely to be visible and appears to target only images created by free users. The code explicitly references “Image-Gen-Watermark-for-free,” suggesting that OpenAI intends to apply such marks solely to outputs from non-paying users. These watermarks cannot be removed by the user, whereas images generated by paid subscribers may remain watermark-free—potentially serving as an incentive to upgrade to a paid plan.
Nonetheless, these developments currently exist only in the codebase, and OpenAI has not officially announced any changes. It remains to be seen whether the watermarking functionality will ultimately be deployed, and in what form.
Related Posts:
- Google’s SynthID Now in Magic Editor: AI Image Detection
- OpenAI announces a bug bounty program, providing a bug bounty ranging from $200 to $20,000
- OpenAI Considers Ads for ChatGPT: Will Free Users Pay the Price?
About The Author
