What is Google Imagen AI? How to try the Beta version?

What sets Imagen AI apart is that its initial release has very specific limitations. You can use AI models to create buildings with different themes or animate creatures.

Here's what you need to know about Imagen AI and how to try it out in its beta.

What is Google Imagen AI?

Imagen AI is a text-to-image model developed by Google. It's an interesting addition to the growing list of AI text-to-image tools, such as DALL-E 2, Dream by Wombo, and Stable Diffusion, all of which can generate visuals great from the written description content.

Google released Imagen through an app called AI Test Kitchen. This is where Google likes to test various AI projects before they are released to the public.

 

The dataset used to train Imagen is LAION-400M. Not all AI companies choose to make this information public, and DALL-E 2 is perhaps the best example of this.

The use of datasets has also become a subject of controversy, in part because it collects millions of images from the Internet. In particular, artists have criticized AI companies because they have not received the consent of the owners of the images or artwork before using them to train their AI models.

If you're worried about your image, you can find out if your image is being used to train an AI model and opt out.

With this in mind, Google cautiously released Imagen. For the first time through beta access, you can try through the AI ​​Test Kitchen app. From the first collection of images created by Imagen, it can be seen that its power lies in producing realistic outputs. You can visit the Imagen research site to see the images for yourself.

How is Google Imagen AI different from DALL-E or Midjourney?

What makes Imagen different is that it is designed to produce two very specific outputs based on functions called City Dreamer and Wobble.

City Dreamer is somewhat reminiscent of the Sim City game, where you build buildings to create a town. With Imagen, you can describe something like a house made of S'mores cakes and your description will be turned into an image.

On the other hand, Wobble creates unique creatures based on your description and they look like animated Pixar characters (like Monsters Inc.). Some things you can try are what kind of clothes the creature will wear, or what material the creature is made of.

On a technical level, Google's research into AI text-to-image systems has found that using larger language models is key to producing more relevant, high-quality images. with text description.

However, since you're limited to creating a single building or creature, it's difficult to compare Imagen to things like DALL-E or Stable Diffusion, at least for now. If you haven't tried it, see how to use DALL-E to create images from text.

 

How to try Imagen Beta

Currently, Imagen is only available to a small group of people during its beta release, via the AI ​​Test Kitchen app. It gives Google a valuable opportunity to get feedback from users and fix any issues with the model before it goes mainstream.

If you want to try Google Imagen AI for yourself, you can sign up for the Imagen AI beta on the AI ​​Test Kitchen website. During the registration process, you will be asked to state your country, device (Android or iOS), occupation, and why you want to use AI Test Kitchen.

If you are one of the select few who got the chance to try Imagen, be sure to provide plenty of feedback. While there are some obvious pros and cons to creating artwork with AI, hopefully in the future, AI models will be safe for everyone to use.

It won't be long before big tech companies like Google start experimenting with AI models. Imagen is a text-to-image AI tool that is sure to be fun to use. Sign up and download Google's AI Test Kitchen app to see more projects in development before anyone else!