Imagen 2 can generate emblems, lettermarks, and abstract logos, and overlay these onto various surfaces. It can also understand more descriptive, long-form prompts and provide detailed answers to questions about elements in an image. Despite concerns over AI-generated disinformation, Google has not revealed the data used to train Imagen 2, and does not offer an opt-out mechanism or creator compensation. Instead, it offers an indemnification policy to protect eligible Vertex AI customers from copyright claims related to Google’s use of training data and Imagen 2 outputs.
Key takeaways:
- Google has made the second generation of its AI model, Imagen 2, more widely available to Google Cloud customers using Vertex AI. The model can create and edit images based on a text prompt.
- Imagen 2 has improved image quality and new capabilities such as rendering text and logos. It can also render text in multiple languages and overlay logos on existing images.
- Google has not disclosed the data used to train Imagen 2 and does not offer an opt-out mechanism or compensation for creators who may have contributed to the data set.
- Google offers an indemnification policy that protects eligible Vertex AI customers from copyright claims related to the use of training data and Imagen 2 outputs, but creators do not have any protection or compensation.