- Google introduces Imagen 2, offering image editing and short videos from text prompts.
- The low-resolution videos target marketers but lag behind competitors.
- Google’s indemnification policy does not cover the preview feature.
After launching in preview, image editing with Imagen 2 is now generally available in Vertex AI, along with two new capabilities: inpainting and outpainting.
These features, already offered by popular image generators like DALL-E, allow users to remove unwanted parts of an image, add new components, and expand the borders of an image to create a wider field of view.
Introducing text-to-live images
The real highlight of the Imagen 2 upgrade is “text-to-live images,” which enables the creation of short, four-second videos from text prompts.
Google is marketing live images as a tool for marketers and creatives, such as generating GIFs for ads featuring nature, food, and animals.
Although the current resolution is low at 360 pixels by 640 pixels, Google promises future improvements. To address concerns about deepfakes, Imagen 2 will employ SynthID, an approach developed by Google DeepMind, to apply invisible, cryptographic watermarks to live images.
Room for improvement
While Google emphasizes the safety and security measures in place for live images, the current offering falls short compared to existing video generation tools like Runway, Stable Video Diffusion, and OpenAI’s Sora.
Google’s own experiments, such as Imagen Video and Phenaki, demonstrate more impressive capabilities. The lack of transparency regarding the training data used for Imagen 2 and the absence of an opt-out tool or compensation for creators whose work might have been used in the model training process raise concerns.
Additionally, text-to-live images is not covered by Google’s generative AI indemnification policy, as it is still in preview.
