Google Bard chatbot now works on Gemini Pro and supports Imagen 2

Google Bard chatbot now works on Gemini Pro and supports Imagen 2

Google has updated the Bard chatbot. It has received support for the large language model Gemini Pro, as well as the Imagen 2 model for imaging.

Gemini Pro was added to Bard for English-speaking users back in December, but the model is now available in more than 40 additional languages ​​in 230 countries and territories. Google has also expanded its double-check feature to 40 languages, allowing users to cross-check answers by automatically searching the web to confirm or refute a chatbot’s claim.

In addition, Bard can now also generate images with text descriptions in more than 40 languages. The chatbot was supported by the Imagen 2 model. It was trained on higher-quality pairs of images and descriptions so that the model could generate more detailed images that better match the semantics of language prompts. Bard Imagen 2 is available for free unlike ChatGPT Plus which requires a paid subscription to generate photorealistic images in various styles.

Google said Bard’s imaging capabilities were “designed with responsibility in mind.” The company’s SynthID watermarking feature will be embedded in the pixels of images, allowing them to prove their origin. Google has implemented other security measures to avoid the creation of images of famous people and to limit the generation of violent, offensive or sexualized content.

The company also released an experimental tool for working with photos based on Imagen 2 called ImageFX. It allows users to create images with simple text prompts, and feedback on the work will be sent directly to Google engineers. As with Bard, all photos created with ImageFX will be tagged with a SynthID and must comply with Safe AI principles.

In late 2023, Google unveiled Imagen 2, a model that will generate more realistic images using a natural distribution of training data instead of a pre-programmed style. It has improved context understanding, renders realistic hands and human faces, and can expand the source image.

In January, the developers announced that they will add an image generation function to the Bard chatbot.

Related posts