Google DeepMind this morning announced two more new generative artificial intelligence (AI) models, namely Veo 2 which can generate video and Imagen 3 which can generate images. Veo 2 can generate videos up to 2 minutes long at 4K resolution. To prevent misuse, all videos generated by Veo 2 will have a SynthID watermark that is invisible to humans but can be detected by computers.
Among the improvements to Veo 2 is the elimination of video issues that look strange such as extra fingers and objects that look absurd. Imagen 3 can also generate images that look more realistic in various styles compared to previous Imagen models.
Veo 2 is already available to selected users through the VideoFX service. Other users will have to enter a waiting list to try it. Meanwhile, Imagen 3 is available through ImageFX in 100 countries starting today.