OpenAI May Soon Add Watermarks for Image Generation Models

OpenAI is reportedly testing a new "watermark" feature for its image-generating model, which is part of the ChatGPT 4.0 suite. This move is particularly intriguing, especially as more users are creating artwork in the style of Studio Ghibli using the Image Gen model.

Not only can the model generate images that include text, but it also produces realistic visuals reminiscent of the renowned Japanese studio's art. With this latest addition, ChatGPT has seen a significant boost in popularity.

Initially available only to paid users, the ChatGPT ImageGen Model is now accessible to everyone, albeit with some limitations. It's worth noting that the watermark feature seems to be aimed specifically at free-tier users. However, it's still uncertain whether OpenAI will proceed with implementing the watermark. The company hasn't confirmed this plan, and changes could be on the horizon.

In a previous blog post, OpenAI mentioned that they trained the model on a vast array of images and texts, which is why the generated visuals are not only aesthetically pleasing but also practical.

"We trained our models on the joint distribution of online images and text, learning not just how images relate to language, but also how they relate to each other," OpenAI noted.

"Combined with aggressive post-training, the resulting model has surprising visual fluency, capable of generating images that are useful, consistent, and context-aware."

OpenAI has also confirmed that it is developing an ImageGen API that will enable developers to create their own products.