Google has unveiled several new features for its video AI model Veo 2, designed to enhance users' ability to create cinematic-quality videos and edit real-world footage. These new capabilities are now available for preview through Google Cloud's Vertex AI platform. Additionally, Google has upgraded its text-to-image generator Imagen 3 and various audio-related AI models.

Veo 2's new features include video inpainting and outpainting. Video inpainting can automatically remove background images, logos, or distracting elements from videos, while video outpainting extends original video frames into different formats, filling in with AI-generated content that blends seamlessly, similar to Adobe’s image extension tools.

Furthermore, Veo 2 users can now select preset cinematographic techniques when generating videos, such as time-lapse, drone perspectives, and simulated panning in multiple directions to guide shot composition, angles, and pacing. A new interpolation feature allows for smooth transitions between two static images by automatically generating intermediate frames for the start and end sequences.

For the text-to-image model Imagen 3, Google has enhanced its editing capabilities, significantly improving object removal results for a more natural look after eliminating unwanted elements. Both Veo 2 and Imagen 3 are already being used by companies like L'Oréal and Kraft Heinz to streamline the production of marketing content, drastically reducing creation timelines.

In the audio domain, Google has launched the private preview of its text-to-music model Lyria and introduced the "instant custom voice" feature for its synthetic voice model Chirp 3. With just 10 seconds of audio input, Chirp 3 can now generate highly realistic custom voices. It also adds a call transcription function that identifies and separates different speakers for clearer records.

Beyond these updates, Google announced several other AI-related advancements. The efficiency-optimized Flash model Gemini 2.5 Flash is set to launch on Vertex AI soon, which dynamically adjusts processing times based on task complexity to accelerate responses for simpler requests. Google has also updated its enterprise-focused Agentic AI tools, enabling AI agents to communicate and perform tasks across platforms like PayPal and Salesforce. Meanwhile, Google Cloud Marketplace now features a dedicated section for purchasing third-party AI agents, making it easier for businesses to explore and buy solutions.