Google is making significant strides in the realm of artificial intelligence by enhancing its video AI model, Veo 2, aiming to empower users to create cinematic-quality videos with ease. The newly introduced capabilities are currently available for preview through Google Clouds Vertex AI platform. This update comes alongside other improvements to Googles text-to-image generator, Imagen 3, as well as enhancements to its audio AI models.

Among the standout features of the updated Veo 2 is the inpainting tool, which allows users to automatically eliminate unwanted elements from their videos. Whether its a distracting logo, an unsightly background image, or any other diversion, Google claims that this feature can seamlessly clean up video content. Additionally, the outpainting functionality allows users to extend the original video frame into a different format. This innovative feature generates new footage that integrates smoothly with the existing video, reminiscent of Adobes Generative Expand feature used for images.

When users stretch their videos beyond their original dimensions using Veo 2, the empty spaces are filled with AI-generated footage that matches the aesthetic and narrative flow of the video. To illustrate this functionality, Google has provided a GIF demonstrating how Veo 2 accomplishes this task.

Moreover, the update introduces cinematic technique presets, which users can select to accompany their video generation. These presets aid in determining shot composition, camera angles, and pacing in the final product. Some examples of these presets include timelapse effects, aerial drone-style perspectives, and simulated camera panning in various directions.

Another noteworthy addition is the new interpolation feature, which enables users to create smooth video transitions between two still images. By simply specifying a starting and ending point, Veo 2 can generate a video that effectively bridges the two images with seamless motion.

Adobe is a notable competitor in this space, with its Firefly video model offering similar capabilities. Just last week, Adobe launched a generative AI video extending feature in Premiere Pro, which underscores the competitive landscape. In terms of digital attribution, Google has incorporated SynthID watermarks into its AI-generated outputs, akin to Adobes Content Credentials system. However, Adobe goes further by ensuring its tools are commercially safe, as they are trained on licensed and public domain contentsomething Google cannot fully guarantee given its approach of training AI models on a vast array of web data.

In addition to the enhancements to Veo 2, Google has also updated its text-to-image model, Imagen 3, to make significant advancements in automatic object removal. According to the company, these improvements lead to more natural outcomes when distractions are removed from images, minimizing any warping of surrounding features. Esteemed brands such as LOreal and Kraft Heinz are already leveraging both Veo 2 and Imagen 3 for their marketing content production, with Kraft Heinzs digital experience leader, Justin Thomas, stating that tasks that previously took them eight weeks can now be completed in just eight hours.

On the audio front, Google has debuted its text-to-music model, Lyria, which is currently available in private preview. Furthermore, an innovative Instant Custom Voice feature has been rolled out for its synthetic speech model, Chirp 3. This upgrade allows Chirp 3 to generate realistic custom voices based on just 10 seconds of audio input. Additionally, a new transcription feature is being launched in preview mode that can effectively identify and separate different speakers, which enhances the clarity of transcriptions during conversations involving multiple participants.

The aforementioned updates represent just a fraction of the AI-related enhancements Google announced recently. The latest iteration, Gemini 2.5 Flash, is set to debut on Vertex AI soon. Google has highlighted that Gemini 2.5 Flash will automatically adjust processing times based on task complexity, ensuring quicker results for simpler requests.

Furthermore, this week, Google is also upgrading its enterprise-focused Agentic AI tools, enabling AI agents to communicate inter-platform, including on services like PayPal and Salesforce. In an effort to streamline access, Google is launching a new section within its Cloud Marketplace, allowing companies to explore and acquire AI agents developed by third-party Google partners.