In partnership with
n
Google just taught its AI to sing like a human.
Lyria 3.5 dropped yesterday with vocals so real, you'll forget a machine made them.
This isn't background music anymore it's the future of songwriting.
But wait, there's more:
HeyGen's new tool turns your Google Doc into a full AI podcast with two hosts
Google's robots can now walk, balance, and grab things all at once
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
• Google's Lyria 3.5 makes AI singers sound almost human
• HeyGen turns your docs into full AI video podcasts
• Google's new robot brain controls the whole body at once
Google's New AI Music Model Sounds Shockingly Real
Google just fired a massive shot at AI music rivals with a brand new model.
DeepMind dropped Lyria 3.5 on July 29, 2026, bringing crazy realistic vocals to its Flow Music platform.
You can finally tweak specific parts of a song without starting over.
The details:
Lyria 3.5 generates songs up to 3 minutes and 4 seconds long across eight languages.
The AI vocals now pack way more emotion and the pronunciation is much clearer.
A new tool called Selective Section Painting lets you fix a bad chorus without ruining a perfect intro.
New users get 500 free credits to test the model right now on the Google Flow Music website.
Google stayed totally quiet about copyright rules and SynthID watermarks for this specific launch.
Why it matters:
AI music is moving past cheap tricks.
Google wants to steal the crown from giants like Suno by giving you real producer tools.
Letting users edit specific chunks of a song changes the game entirely.
It turns random generation into actual music production.
💡 L8R's Take:
They are obviously dodging the hard questions about their training data.
Play with it for fun, but do not build a business on Flow Music just yet.
From Our Partner
One idea shouldn't take six rewrites to post.
Posting everywhere means rewriting one idea six times, so you post to one, or none. SureThing turns one idea into native posts for every platform.
Two AI Hosts, One Click: HeyGen's New Podcast Tool
You can now turn a boring PDF into a full video podcast with two hosts.
HeyGen just dropped a wild new tool that automates your whole studio setup.
It builds a shared scene where two AI avatars talk to each other like real humans.
The details:
Users paste a web link, upload a PDF, or type a topic to start generating the show.
The AI creates a real shared scene, putting two avatars together instead of gluing separate clips side-by-side.
It handles all the heavy lifting, including multi-camera wide shots, close-ups, and B-roll footage.
You can invite a coworker via email to be your AI co-host, even if they do not have a HeyGen account.
If you change a line in the script before publishing, the AI instantly updates the voice and lip-syncing.
Why it matters:
Making good video content usually takes cameras, lights, and hours of editing.
This tool wipes all that out. Corporate teams and creators can now recycle old blog posts or dry newsletters into fresh YouTube videos in minutes.
It lowers the barrier to entry for high-quality video production to almost zero.
💡 L8R's Take:
People are arguing online about whether AI podcasts feel authentic, but they are missing the point.
Viewers care about getting good information, not how the sausage gets made.
If a fake podcast explains a boring company report better than a real human can, the AI wins.
From Our Partners
Your prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
Google DeepMind Gives Robots a Full-Body Brain
Robots just got a massive brain upgrade that controls them from head to toe.
Google DeepMind dropped Gemini Robotics 2 to power everything from walking feet to grasping fingers.
This changes the game for humanoid robots by giving them one smart system to handle their whole body.
The details:
The lineup features three models, including an Embodied Reasoning model that lets robots plan complex tasks and work together as a team.
A lightweight On-Device model works without the internet and learns to control a brand new robot body in a few hours using fewer than 200 examples.
Google proved the tech works by running the AI on hardware from Boston Dynamics, Apptronik, and Franka.
A new safety feature makes the robots freeze instantly if a human gets too close to them.
Why it matters:
We are moving past the days of training a robot for just one boring task.
This system gives one AI the power to run many different types of robots.
It is a huge step toward physical AGI, where machines can finally handle complex real-world jobs like warehouse work or home chores.
💡 L8R's Take:
Google is playing this perfectly by building the brain and letting other companies build the metal bodies.
This makes them the ultimate puppet master of the robot world.
Still, their warning to keep this out of healthcare and transport shows they know the tech is not foolproof yet.
🚀 Quick L8R Summary
Google Lyria 3.5: Google's music AI now sings like a real human better emotion, clearer words, smarter melodies.
HeyGen Video Podcast: Turn any idea or link into a full AI-generated podcast with two hosts, cameras, and B-roll in minutes.
Gemini Robotics 2: Google's new AI gives robots full-body control they can walk, balance, and grab things all at once.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋