OpenAI's Secret AI Image Generation Tool Set to Debut

2023-08-02

OpenAI has been silent since its last release, but now it is secretly developing its next AI image generation tool. In a competition to create realistic AI images, one major participant who gained public attention with its text-to-image model seems to have remained silent since last year. While OpenAI has been busy with ChatGPT, other participants like Midjourney and Stable Diffusion have surpassed OpenAI's image generation platform, Dall E-2. However, according to the latest developments, Dall E-3 is on its way to catch up in the AI image generation race. It is believed that OpenAI is testing a new image generation platform, possibly an upgraded version of Dall E-2. Through an exclusive preview limited to 400 invitees, the latest version of this model can be accessed on OpenAI's dedicated test servers. Youtuber MattVidPro shared images of the new model being tested in an explanatory video. Users' feedback has been positive, with comments like "After using this, I have no interest in Midjourney." The new model is said to be powerful and advanced in following prompts and producing coherent details. It can create images with detailed features such as hair, lighting, ad copy, and even addresses common issues with hand details. It has been compared to other applications like Midjourney V5.2 and Stable Diffusion XL, and seems to outperform them. After removing the waiting list for users, Dall E-2 was released to the public in September 2022. Since then, the model has not received any major updates. Reports in March this year suggested that the company was experimenting with Dall E-2 and seeking early feedback from a small number of users. The model was used to create clearer and more realistic images. With GPT-4 having multimodal capabilities, OpenAI's next text-to-image generation model may have even more powerful features. Midjourney has released five versions of its text-to-image generation model in the past year, consistently using closed-source models. On the other hand, Stability Diffusion is open-source, and their latest model, Stability Diffusion XL 1.0, can also be used on Amazon Bedrock. Adobe Firefly challenges Midjourney and Dall E with its generative AI capabilities, offering trial services followed by subscription options. OpenAI recently pledged to take a series of actions to ensure responsible AI governance. In coordination with the US government, OpenAI, along with six major tech companies including Microsoft, Google, and Meta, will work on adding watermarks to AI-generated audio and visual content. This may be embedded in their latest tested versions. If so, OpenAI will become the first major tech company to label AI-generated images. While safety seems to be their top priority, the current model still has a long way to go in terms of safety. Due to being in the testing phase, the current model lacks safety features and can generate images containing blood, violence, and explicit content. It may also pop up images depicting extreme violence without any prompts. Additionally, it can generate copyrighted artworks, characters, and accurate company logos. Last year, Dall E-2 faced scrutiny for creating inappropriate images, reportedly reinforcing gender bias and racial discrimination. Although the new model needs fine-tuning and refinement to introduce safety features, the community's feedback on this model has been very optimistic, rating it higher than current image generation tools. The new model is expected to be launched in December. Source: https://analyticsindiamag.com/openais-secret-image-generation-tool-to-debut-soon/

RECENT AI NEWS

RECENT AI TOOLS