Google Imagen 3 is the company's current best AI image generation model. It was first introduced at the I/O event in May but wasn't available until June. Even so, only limited Vertex AIusers were able to use it, but now, the company has quietly opened access to all US users.

In the Google I/O event of May, Google said that this text-to-image AI model could offer "better detail, richer lighting, and fewer distracting artifacts" compared to the previous AI image generation models. Per the experiments done by Reddit users, it seems that the company did deliver these promises, at least to some extent.

According to the early testers, the images generated by Imagen 3 can offer better texture and have improved word recognition. The AI image generator also provides an option to edit photos. Similar to Magic Editor on Google Pixel (pre-order the Pixel 9 on Amazon), you can highlight a specific part of the picture and write what change you want in that region.

The company has also published a research paperdetailing the public launch and the Imagen 3 model. This launch came right after the introduction of Grok-2, a rivaling image generator by Elon Musk's xAI. However, compared to Grok-2, Google's latest model has a couple of restrictions in place.

For example, it doesn't allow you to generate images of public figures, nor will it create inappropriate content. These are likely in place to avoid controversies. As for accessing the AI image generator, you can use it through ImageFX if you're in the US. For others, DALL-E 3 in ChatGPT and Microsoft Designer Image Creator are good free alternatives.