Lyria 3 in Gemini: AI Music & Major Moves

Good morning, {{ first_name | AI enthusiasts }}

AI music generators have already produced tracks that can fool casual listeners—if you knew where to look. Now Google has rolled Lyria 3 into Gemini, making AI music creation accessible to millions. Read on for today's AI rundown, including the biggest product moves, new tools, and community workflows.

Share your own AI use case for a chance to appear in the Community AI Workflows section below.

In today's AI rundown

• Google brings AI music creation into Gemini
• OpenAI hires Instagram's Hollywood dealmaker
• Turn product photos into scroll-stopping videos
• Tavus' new AI avatars actually read the room
• 4 new AI tools, community workflows, and more

GOOGLE: Google brings AI music creation into Gemini

The Rundown

Google launched Lyria 3 inside Gemini, a new AI music model that turns a text prompt, photo, or video into a customized 30-second track. Each output includes auto-generated lyrics and cover art.

The details

• Lyria 3 manages genre, tempo, vocal style, and lyrics automatically. Users can start from text, photos, or videos.
• DeepMind developed Lyria since 2023; this is its first broad consumer release through Gemini.
• All tracks get Google's SynthID watermark. Gemini also lets users upload audio to check if it was .
• YouTube creators can access Lyria 3 via Dream Track for Shorts to add custom audio to short-form videos.

Why it matters

AI music platforms such as Suno and Udio were already producing convincing tracks for niche users. Embedding Lyria in Gemini changes distribution: millions of mainstream users can now create music from a prompt. That matters for creators, marketers, and anyone making short video content or gaming news montages that need royalty-free tracks.

TOGETHER WITH VISA: Meet Visa Intelligent Commerce

The framework enables AI agents to find and buy while providing tools and safeguards for secure agentic payments.

• Tokenized credentials for secure payments
• Spend controls and strong user authentication
• Controlled GenAI-assisted shopping on your terms

OPENAI: OpenAI poaches Instagram's Hollywood dealmaker

The Rundown

OpenAI hired Charles Porch, Meta's longtime celebrity partnerships chief, as VP of global creative partnerships. The move aims to bridge OpenAI with a skeptical entertainment industry.

The details

• Porch has 15+ years building celebrity programs on Instagram, from major album drops to high-profile religious outreach.
• The hire follows OpenAI's $1B Disney deal to open animated characters on its Sora video platform.
• Porch plans a "listening tour" across creative communities this spring and reports to OpenAI's applications CEO Fidji Simo.

Why it matters

Hollywood remains wary of AI. Porch's relationships and track record make him uniquely qualified to translate between talent and tech. That could shift sentiment across studios, agencies, and high-profile creators—impacting everything from branded content to licensing deals tied to gaming news and entertainment coverage.

AI TRAINING: Turn product photos into scroll-stopping videos

The Rundown

This practical guide shows how to convert a product photo into a short cinematic clip for social media using Runway's video generation tools. The workflow creates consistent, ad-ready clips for campaigns and product announcements.

Step-by-step

1. Grab a product photo. If you don't have one, use a software mockup or ask Gemini to generate a brand-appropriate image.
2. Drop the photo into any AI and prompt: "Write me a brief video prompt to use in Runway to generate a social media video based on this product photo. Add animations and/or camera movement if applicable."
3. Open Runway > Tool > Video, add your starting frame, paste the prompt, and generate the clip.
4. Repeat for other product photos. Test different text overlays and music when posting to social channels.

Pro tip: Add these clips to Google or Meta ad campaigns to increase engagement and conversion.

PRESENTED BY VANTA: Compliance just got 82% faster

Vanta's AI-powered platform streamlines compliance, risk, and customer trust—helping companies automate SOC 2, ISO 27001, HIPAA, and more.

• Automate up to 90% of audit work
• Use AI-powered questionnaires to speed security reviews
• Spend up to 82% less time on audits (customers include Ramp and Writer)

RESEARCH: Tavus' new AI avatars actually read the room

The Rundown

Tavus introduced Phoenix-4, a real-time human rendering model that creates AI avatars with full facial expressions and emotional transitions during live conversation.

The details

• Phoenix-4 generates every facial pixel per frame and was trained on thousands of hours of human conversations.
• It supports 10+ emotional states and switches smoothly between them to avoid awkward or mismatched reactions.
• Runs at HD resolution and ~40 FPS for smooth, live interactions.
• Targeted use cases: healthcare, education, and sales—where perceived empathy or attentive listening can change outcomes.

Why it matters

Text-based AI is already highly conversational; video avatars are catching up. More human reactions improve user trust and engagement, which benefits legitimate applications. At the same time, these advances raise risks for deception and misuse.

Trending AI Tools

• Unwrap Customer Intelligence — Convert unstructured feedback into product insights
• Lyria 3 — Google's AI music model inside Gemini
• Phoenix-4 — Tavus' real-time emotional avatar AI
• Grok 4.20 — xAI's upgraded model with new agentic features

Everything else in AI today

• Ex-DeepMind researcher David Silver is reportedly raising $1B at a $4B valuation for Ineffable Intelligence—Europe's largest seed round.
• Fei-Fei Li's World Labs announced a $1B round including $200M from Autodesk to expand world models into 3D and entertainment.
• Perplexity may be removing ads to protect perceived neutrality in AI answers.
• OpenAI acqui-hired enterprise AI search startup Nerve to boost ChatGPT's search capabilities.
• The trailer for "The AI Doc: Or How I Became an Apocaloptimist" is gaining attention with interviews from major AI leaders.

Community AI workflows

Every newsletter, we highlight a reader workflow that saves time or solves a real problem. Today's submission comes from Pradeep A. in Washington, DC.

"My garage door wasn't closing. I asked Claude…and it asked me to look at the sensor lights. Upon checking, I discovered there was no light in one of the sensors. It asked me a couple of questions about the make of the motor, and then gave me the part number for the sensor light. I got the part from Amazon and installed it myself without even having to call a mechanic."

How do you use AI? Tell us here.

See you soon,
Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown