Google Cloud brings music generation to Vertex AI with Lyria and upgrades Veo 2

If you have ever tried building an AI-generated video ad from scratch, you know the headache: generate stills in one app, animate them in a second, clone voices in a third, and hunt down background music in a fourth.

Google Cloud just consolidated that entire pipeline. The company added Lyria, its text-to-music model, to Vertex AI in preview with an allowlist. With that move, Vertex AI becomes the only enterprise cloud platform packing generative models across video, image, speech, and music under one roof.

The complete creative stack

Lyria is designed to spit out high-fidelity music tracks straight from descriptive text prompts. Google is pitching it directly at marketing teams, video editors, and podcasters looking to bypass expensive royalty-free licensing libraries.

Alongside the music release, Google rolled out hefty editing upgrades to Veo 2, its video generation model. Teams can now do inpainting to cleanly erase unwanted logos or background clutter across every frame, outpainting to convert horizontal landscape videos into vertical social clips, and video interpolation to smoothly stitch two separate assets together.

Audio and image models got tune-ups as well. Chirp 3 gained Instant Custom Voice, which clones a realistic voice using just 10 seconds of source audio, plus new speaker diarization for transcribing multi-person meetings. Meanwhile, Imagen 3 received upgraded inpainting and object removal capabilities.

Google wrapped all of this in its enterprise shield: SynthID invisible watermarking across images, video, and audio, a pledge not to train on customer data, and intellectual property copyright indemnification for commercial deployments. Enterprise heavyweights like WPP, Agoda, and L'Oreal Groupe are already tapping the models in production.

Why it matters

For enterprise marketing and creative teams, running an end-to-end multimedia asset pipeline through a single API dashboard is a massive operational win. Instead of juggling multiple vendor subscriptions and clearing copyright risks with four different legal departments, Google provides the entire creative suite with a built-in legal shield.

That said, access remains tightly guarded. Lyria, Veo 2's advanced editing tools, and Chirp 3's 10-second voice cloner are all locked behind preview allowlists. Enterprises will still have to go through compliance checks before putting custom voices into production.

We think this is a smart defensive play by Google Cloud. Rather than competing model-by-model against specialized startups in video or audio, Google is leveraging its enterprise distribution muscle. Single-modality AI startups will face increasing pressure to prove why customers should pay for a standalone tool when their cloud provider can bundle the whole production studio.