You can now give Google’s AI video model camera directions

Google is trying to marque it easier for users of its video AI exemplary Veo 2 to marque cinematic-looking generations and edit existent footage. The new Veo 2 capabilities are disposable to preview via Google Cloud’s Vertex AI platform, alongside different updates to amended Google’s text-to-image generator, Imagen 3, and audio-related AI models.

New Veo 2 features see inpainting, which tin automatically region “unwanted inheritance images, logos, oregon distractions from your videos” according to Google, and outpainting, which extends the framework of the archetypal video into a antithetic format. The second instrumentality volition capable the caller abstraction with ai-generated video footage that blends into the archetypal clip, akin to Adobe’s Generative Expand feature for images.

The update besides lets Veo 2 users prime cinematic method presets to see alongside their substance descriptions erstwhile generating footage, which tin beryllium utilized to assistance usher changeable composition, camera angles, and pacing successful the last results. Example presets see timelapse effects, drone-style POV, and simulating camera-panning successful antithetic directions.

A caller interpolation diagnostic has besides been added that tin make a video modulation betwixt 2 inactive images, filling successful the opening and extremity sequences with caller frames.

Adobe’s competing Firefly video exemplary has immoderate akin capabilities, with a generative AI video extending diagnostic launching successful Premiere Pro past week. Google besides adds SynthID integer attribution watermarks into its AI-generated outputs, overmuch similar Adobe’s Content Credentials system, but Adobe goes a measurement further by pledging that its tools are afloat commercially harmless due to the fact that they’re trained connected licensed and nationalist domain contented — thing Google can’t lucifer aft inhaling the web to bid its AI models.

Editing capabilities successful Google’s text-to-image exemplary Imagen 3 person besides been updated to “significantly” amended automatic entity removal, according to Google, providing what are expected to beryllium much earthy results erstwhile removing distractions. Both Veo 2 and Imagen 3 are already being utilized by companies similar L’Oreal and Kraft Heinz for selling contented production, with Kraft Heinz’s integer acquisition person Justin Thomas saying the benignant of task that “once took america 8 weeks is present lone taking 8 hours.”

On the audio side, Google has released its text-to-music model, Lyria, successful a backstage preview and rolled retired an “Instant Custom Voice” diagnostic for its synthetic code model, Chirp 3. Google says that Chirp 3 tin present make “realistic customized voices from 10 seconds of audio input,” and that a caller transcription diagnostic is launching successful preview that tin place and abstracted idiosyncratic speakers to supply clearer transcriptions for calls wherever aggregate radical are talking.

These updates are conscionable a fistful of AI-related announcements that Google made today. Gemini 2.5 Flash, the latest mentation of the company’s efficiency-optimized Flash model, volition soon beryllium disposable connected Vertex AI. Google says that Gemini 2.5 Flash “automatically adjusts processing time” based connected the complexity of the task to supply faster results for elemental requests.

Google is besides updating its enterprise-focused Agentic AI tools this week to let AI agents to pass with each other and execute tasks crossed platforms similar PayPal and Salesforce. Meanwhile, a caller conception is being launched connected Google’s Cloud Marketplace for companies to browse and acquisition AI agents built by third-party Google partners.