Google released Veo 3, its state-of-the-art video generation model, in May this year. Then in July, the company utilized the power of Veo 3 to add photo-to-video capability to its Gemini generative AI model. Now Google has announced Veo 3.1, heralding it as the next leap in generative video.
Veo 3.1
Google says that Veo 3.1 builds on Veo 3, with stronger prompt adherence. This means that the AI should follow your text prompts and how the words relate to each other more closely. Veo 3.1 also promises richer audio, more narrative control, and enhanced realism that captures true-to-life textures.
In addition, Google says that Veo 3.1 delivers improved audiovisual quality when turning images into videos.
Reference images
With Veo 3.1, you can now guide the AI video generation process by providing up to three reference images of a character, object, or scene. Google says that this is helpful when you want to maintain character consistency across multiple shots or apply a specific style to your video.
Scene extension
The scene extension feature with Veo 3.1 means that your story is no longer limited by the original video generation. With scene extension, you can generate new clips that connect to your previous video. Veo 3.1 generates each new video based on the final second of the previous clip. As a result, you can create much longer videos, lasting as long as a minute or even more.
First and last frame
With Veo 3.1, you can also create videos by specifying the first and last frame. Google says that Veo 3.1 will then generate smooth, natural scenes that bridge your two images. As well as creating a transition between the images, Veo 3.1 will create the accompanying audio.
Add new element
Using Veo 3.1 in Flow, you can now add new elements to any scene. Google says that with “Insert”, you can introduce anything from realistic details to fantastical creatures. Flow will also now handle complex details like shadows and scene lighting so that your additions look natural. Also, you will soon be able to remove objects as well, with Flow reconstructing the background.
What we think
The updates and improvements in generative video models seem to be coming faster and faster. Google’s Veo 3.1 updates are undeniably impressive, and the company is working hard to give users more control and consistency in their generated videos. The output is stunning at times as well, with animated or fantasy content.
However, there is still something of the uncanny valley about AI’s attempts to generate realistic human representations. There is also a wider ethical debate about the training and use of generative AI which needs to be resolved.
Availability
Google has released Veo 3.1 and Veo 3.1 Fast in paid preview in the Gemini API. These new models are available via the Gemini API in Google AI Studio and Vertex AI. Veo 3.1 is also available in the Gemini app and Flow.
