Google is rolling out the Imagen 3 AI image generation model to users, adding improved capabilities

Google released its in-house image-generating artificial intelligence (AI) model, Imagen 3, on Thursday. The tech giant did not announce the release, instead rolling out the model to users quietly. Additionally, a research paper detailing how the image-generating model works was also published in an online journal. The text-to-image model is currently only available to users in the US, and there is no word on when it will be made available to users in other regions.

AI Imagen 3 model released by Google

The tech giant’s AI Test Kitchen now lets users sign up to the platform and use its AI model to generate images. The third generation of the Imagen model is said to get improved texture generation and word recognition capabilities, as well as stricter adherence to deadlines.

Since the AI ​​model is only available in the US, Gadgets 360 was unable to test the platform. However, a Reddit user claimed that it was able to generate images in a variety of styles, such as Nikon DSLR quality, GoPro style, wide-angle lens, and more. However, the model struggles to generate close-ups with multiple people and underexposed images, which was possible with its predecessor.

Another area where Imagen 3 has problems is with limbs. A user claimed that the model would generate incorrect results when using prompts such as “guy holding a cup of coffee.” The AI ​​would end up generating additional limbs, creating a random limb holding an object, or merging an object with a limb. The image generation model is also said to have very strict censorship on prompts.

Google also published a research paper in the pre-print online journal arXiv. The company emphasized there that it used a latent diffusion model, which is a variant of the diffusion model popularized by Stable Diffusion. The company also added that new methods were used to minimize potential damage using the Imagen 3 model.

It’s worth noting that the free version of the Gemini chatbot can also generate images, but it uses Gemini’s capabilities to do so. Imagen 3 is built on a different architecture and, because its dataset is largely image-based, it’s better trained for AI image generation.