Veo 2 learns inpainting, outpainting, camera movements and more
[17:27 Thu,10.April 2025 by blip]
Google has announced new features for its video AI Veo 2. These not only allow for more control over generation, but also enable the use of the latest AI model (veo-2.0-generate-001) to manipulate video material of all kinds. With this, Google is catching up with other AI providers who have been offering similar features for some time. In our opinion, however, Veo 2 still operates at an unbeatably high level.
Using inpainting, for example, visible safety ropes can be automatically removed from stunt recordings, making complex rotoscoping jobs superfluous. With the new outpainting, on the other hand, given video sequences can be expanded in area, i.e. the framing can be changed by adding suitable image content.
The virtual camera movements can now be specified more precisely in Veo 2 via prompt, as well as image composition, camera angle and tempo.
✨ Sophisticated cinematic techniques
Teams can direct shot composition, camera angles, and pacing with ease - without requiring complex prompting or specialized expertise.
For example, you can use pre-sets to move the camera in different directions, create a timelapse effect,… pic.twitter.com/lRrFLTiDXv
An interpolation function from start to end image also makes it possible to connect two video sequences with each other.
Veo 2 is part of Vertex AI, Google's generative AI environment, along with the newly available music tool Lyria, Chirp 3 for audio and voice generation, and Imagen 3, the latest text-to-image model. As part of Vertex AI, users in this country can also access Veo 2, while this is still not possible via VideoFX.
Veo 2 can calculate videos up to 8 seconds long, but at a price of 50 cents per second. Detailed documentation on the Veo AI video generator can be found here: https://cloud.google.com/vertex-ai/generative-ai/docs/video/generate-videos?hl=de.
;)
;)
;)