Google Just Dropped LYRIA 3: New AI Feature No One Expected

Google has rolled out significant AI updates, including Lyria 3 for generating high-quality music with vocals, Pomelli's AI-powered Photoshoot for product marketing, and Stitch's advanced design agents with native code integration, all emphasizing integrated, multimodal workflows.

• Lyria 3 generates production-quality music with automatic vocals and lyrics, responding to text, image, or video inputs.

• Synth ID watermarks Lyria 3 outputs for attribution, while Lyria Realtime enables interactive, real-time music generation and control.

• Pomelli's Photoshoot offers AI-powered product photography for businesses, and Stitch's new agents integrate design directly into app development workflows.

Google recently launched Lyria 3, its advanced music generation model now integrated into Gemini and YouTube's Dream Track, capable of creating full 30-second tracks with automatic lyrics and vocals from various inputs. Alongside this, Pomelli introduced 'Photoshoot' for AI-powered product photography for businesses, and Stitch expanded its AI design tools with new agents and native code editor integration. These updates highlight Google's shift towards building integrated AI systems that streamline creative workflows across music, visuals, and design.

Lyria 3 Music Generation

• 00:00:24 Google has officially launched Lyria 3, an advanced music generation model integrated into the Gemini app and YouTube's Dream Track. This model allows users to generate 30-second music tracks with automatic lyrics, vocals, and instrumentation using natural language prompts or by uploading an image or video, signifying music as a first-class modality alongside text and vision. Lyria 3 produces production-quality audio at a 48 kHz sample rate and is designed to maintain long-range coherence for complex, multi-layered arrangements generated from scratch.

Lyria 3 Technical & Safety

• 00:03:04 Every track generated by Lyria 3 includes an imperceptible, embedded watermark using Google's Synth ID technology, which remains detectable even if the audio is compressed or re-recorded, addressing copyright and attribution concerns. Furthermore, Google DeepMind introduced Lyria Realtime, a system that generates music in two-second chunks, allowing for live steering and real-time adaptation to user controls with under 2 seconds of latency. Google also offers a Music AI Sandbox for creators to interactively control and transform musical ideas, treating the AI as a collaborative jamming partner.

Pomelli's AI Photoshoot

• 00:07:34 Pomelli, Google's AI marketing tool for small and medium-sized businesses, is rolling out a new 'Photoshoot' feature. This tool enables businesses to upload a product image, choose visual themes and templates, and generate polished, professional marketing images, eliminating the high cost of traditional product photography. Photoshoot integrates with Pomelli's existing 'business DNA profile' to produce ready-to-use assets for social media and campaigns directly within the platform, demonstrating Google's intent to position Pomelli as a serious contender in SMB marketing automation.

Stitch's Design & Code Agents

• 00:09:22 Stitch, Google's AI design tool, has introduced new agents and features to streamline app design and development. The 'Hatter' agent is designed for complex, multi-step design tasks, implying a deeper reasoning capability similar to 'deep design.' Additionally, Stitch now supports app store asset generation, automatically creating store-ready screenshots and icons, and offers native MCP (model context protocol) integration, allowing direct connection to code editors like Cursor and Gemini CLI, significantly reducing friction for designers and developers to pull designs into coding environments.