Google unveils Veo and Imagen 3, new AI powered generative media models | Technology News
At the Google I/O 2024, the Alphabet Inc company introduced several new AI models that can help with different tasks and also brought some improvements to its existing models. Among its latest announcements were the AI models Veo and Imagen 3, which have been developed to help generate videos and images.
Google says Veo can generate videos in 1080p resolution in different cinematic and visual styles with a length of more than a minute. The multimodal AI can capture tones and even render details in long prompts, capture the tone of the scene and understand natural language and visual semantics.
Veo is also familiar with terms like ‘aerial shots of a landscape’ and ‘timelapse’, offering users more control over what they want Veo to generate and can easily create videos with people, animals and objects moving around in a realistic manner. Veo seems ahead of its peers, as OpenAI’s text-to-video model Sora can only generate high-res videos that are up to 60 seconds long.
Google also said that they are inviting creators and filmmakers to experiment and play around with the new model. Veo is available to a handful of creators who use VideoFX, but some features will also be available to YouTube Shorts creators sometime in the future.
The company also unveiled Imagen 3, an updated version of the text-to-image generator. Compared to previous versions, Imagen 3 is capable of producing photorealistic, lifelike images with much less artifacts. Google says it also happens to be the best-ever model for rendering text and that the updated version can also help generate text-based content like personalised birthday messages and title slides in presentations. It is currently available for creators as a private preview in ImageFX and will soon be coming to Vertex AI.
“Imagen 3 better understands natural language, the intent behind your prompt, and incorporates small details from longer prompts. This additional detail helps Imagen 3 master a range of styles. It’s also our best model yet for rendering text, which has been a challenge for image generation models,” Google stated in its official release.
Google also said that it has been collaborating with musicians, songwriters, and producers, including Wyclef Jean and Marc Rebillet, to develop generative music technologies like Lyria. Moreover, Google’s Music AI Sandbox offers tools for creating and transforming music. The company said that its efforts, especially in partnership with YouTube, showcase AI’s potential in music creation.
© IE Online Media Services Pvt Ltd
First uploaded on: 14-05-2024 at 23:39 IST
