Hyderabad: Google on Monday released a new and improved version of its AI video generator Veo and its image generator Imagen. The new version named Veo 2 enhances the capabilities of the older version launched early in 2024. It can now create 4K resolution videos for longer durations of time. The update has just been added within a week, after the introduction of Veo to Vertex AI for its enterprise clients. The new version -- Veo 2-- is available to select creators in the VideoFX platform accessible through Google Labs.
Meanwhile, Imagen 3 has been rolled out in ImageFX to more than 100 countries. The latest version of the AI image generator comes with an enhanced ability to render more diverse art styles with great accuracy and compose images more efficiently.
Google Veo 2: Improvements
Sudhar Pichai, CEO of Google, shared an X post highlighting the new capabilities of the updated version of Veo 2.
More details here https://t.co/bWHjEsWEXs and at https://t.co/qyYdBqPxZZ
— Sundar Pichai (@sundarpichai) December 16, 2024
The AI video generation model is used as a video background generator tool for YouTube Shorts. Google introduced Veo 2 as a "state-of-the-art video model" that can understand the physics of movement and the real world better. It can understand the science of cinematography and can generate a specific genre using various lenses, cinematic effects, and camera movement.
The AI video generation tool can understand complex prompts like "a low-angle tracking shot that glides through the middle of a scene", or "a close-up shot on the face of a scientist looking through her microscope".
Moreover, Veo 2 can generate videos up to 4K resolution with a duration lasting "minutes" in length. Notably, Google did not mention the exact time duration of the AI videos that can be generated. The company said in a blog post that while most video generation models hallucinate unwanted details, such as an extra finger or unexpected objects, the Veo 2 does it "less frequently".
Google has added its native SynthID watermark on AI-generated videos that reduce the misuse of the footage to create Deepfakes and to identify between real videos and AI videos. Google will gradually roll out the latest AI model to enhance its quality and safety before its global release.
Imagen 3: Global Rollout in ImageFX
The updated Imagen 3 image-generation model can now generate brighter, better-composed images, Google claims, adding that it can now render more diverse art styles with greater accuracy, ranging from photorealism to impressionism and from abstract images to anime styles. The company says that the updated model follows prompts more faithfully and produces richer details with better textures.
Google also released Whisk, its newest experiment from Google Labs, based on its latest Imagen 3 model. This tool allows users to input or create images with more precision and remix them to create something uniquely their own, from a "digital plushie" to an "enamel pin" or sticker. It uses Gemini's visual understanding and description capabilities to automatically write detailed captions of your images and feed them into Imagen 3, allowing for an easy use case.
