OpenAI is actively testing a new feature to add watermarks to images generated by its GPT-4o image generation model. Recently, ChatGPT has gained significant attention for its powerful image generation capabilities. The model not only accurately generates images with text but also creates highly realistic visual works, even mimicking the artistic styles of renowned animation studios like Studio Ghibli.
Image Source Note: Image generated by AI, licensed by Midjourney
Initially available only to ChatGPT Plus users, the ImageGen model has now been fully rolled out to all users. AI researcher Tibor Blaho recently discovered that OpenAI seems to be testing a watermark labeled "ImageGen" on images generated by free users. Sources have confirmed that OpenAI has recently started adding watermarks to images generated by free accounts, while ChatGPT Plus subscribers can save watermark-free images.
OpenAI previously emphasized in a blog post that the GPT-4o model's training data included a massive amount of images and text, giving it not only excellent aesthetic qualities but also broad practicality. OpenAI noted that through extensive joint training on online images and text, the model learned not only the relationship between images and language but also the logical relationships between images. Combined with powerful post-training, the model exhibits stunning visual capabilities, generating images that are not only practical and coherent but also possess a good understanding of context.
Furthermore, OpenAI has confirmed it is developing an ImageGen API, allowing developers to build their own applications using this interface in the future. The move to add watermarks for free users may be a crucial step for OpenAI in further regulating the use of its image generation technology, differentiating user benefits, and signifying a deepening of its commercialization strategy.