Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

stumbled onto something genuinely impressive today - real-time interactive AI avatars with 180ms latency let me break down why I think this is one of the more underrated AI products right now 👇 most AI agents are either a chat box or a robotic voice. anam gives them a... face. photorealistic. micro-expressions. no looping, no mouth-dubbing nonsense. the numbers are hard to ignore: → +44% engagement when your agent has a face vs text/voice only → CARA-3 (their latest model) is #1 across ALL avatar benchmarks in 2025 → 33% faster than the next-best competitor → 180ms avg response time, literally faster than a blink → 70+ languages with native voices → plug in your own LLM, your own voice, your own avatar image use cases they're already disrupting: customer support, sales, medical services, language tutoring, skill training and it's actually developer-friendly. clean SDK, LiveKit + Pipecat integrations, no-code option if you don't want to touch the API if you're building AI products or advising companies on AI strategy, this is one to watchshow more

Vyom

11,579 subscribers

268,158 views • 5 months ago •via X (Twitter)

Education Science & Technology

Anya Rossi• Live Now

Private livecam show

29 Comments

Mark Richards5 months ago

That was so wild I had to record it. it literally felt like a natural, bilingual zoom call. (PS. zero affiliation. I hadn't even heard of them until i saw your post)

Dan Benyamin (Æ)5 months ago

The founders are ex-synthesia, wonder how that's working out 😅

Vyom5 months ago

Why did something before? I’m not aware !

Prasad Pilla5 months ago

this is such a well thought website

Yev Marusenko5 months ago

Infrastructure tool or they are consumer app?

Vyom5 months ago

Consumer

Yev Marusenko5 months ago

Oh surprising actually. Thought it’s for devs. But yea this will be more common soon.

Carsten5 months ago

Website design is fantastic at least !!!

AI Mastery Guide5 months ago

44% more engagement just from giving your agent a face is a number that should make every customer support team stop and rethink their setup

sarwech5 months ago

How does it compare to Tavus?

Vyom5 months ago

Never heard of Tavus

sarwech5 months ago

It’s probably the leader in this space

Muhammad Anas5 months ago

Giving AI agents a realistic face & real-time interaction changes the experience completely. Feels way more human than another chatbot window.

Kanhaiya Yadav5 months ago

Crazy 😳

Ivan Flugelman — VVSVS™5 months ago

you should check @Ojin_ai unless this is an…unannounced paid partnership

Habibi ❎ (💰//acc) | 𐤊5 months ago

@yzzztv_official

Jesse Lau5 months ago

First time I have been blown away by one of these.

Ron Shah5 months ago

this site is mad buggy.

Gavin Payne5 months ago

very cool

Ahmed Goda5 months ago

Hey, I think would make a much stronger brand for an avatar platform than generic brandable names

Mo Ayob5 months ago

This is interesting. Do you know if they do digital cloning ?

Percy Kintu5 months ago

Now give them a real phone number plus memory and it’s over

Rhy5 months ago

gave it a try but the outcome fell short of the marketing hype around it

TJ5 months ago

Plugging my boy @interact_studio

NAITIVE5 months ago

:)

Vadim | Sollas5 months ago

Nice soundtrack

NIHAL JANI5 months ago

This is good but if you want your own avatar, new gemini omni isnt bad.its expressive too. Dont know about latency though.

Parag Arora5 months ago

Stumble upon too someday :)

Adam Delehanty5 months ago

“Genuinely” is slop language

Related Videos

Add a face to your voice agent. LiveAvatar by HeyGen is now supported in LiveKit Agents. Add a realtime human avatar to your agent without rebuilding the conversation loop. Your LiveKit agent still owns the room, turn-taking, model orchestration, and voice pipeline. LiveAvatar renders the synchronized face and video stream. Useful for product demos, onboarding, tutoring, and support agents that need a visual layer.

LiveKit

10,885 views • 5 months ago

What does it actually mean to be AI native? There was no clear guide on the internet for how to become AI native so we built the definitive one (60 min masterclass): 1. An AI native org has 3 layers: people for strategy and taste, agents for execution, and a shared context layer that makes the entire company readable to agents. 2. AI eats the middle of your work. You used to spend 80% of your day on execution. Now agents do that. Your job is the bookends: deciding what to do and judging whether it's good enough. 3. Everyone is a manager now. Your output is the output of your agents. If your agents produce garbage, that's on you. You set them up wrong. 4. Using ChatGPT doesn't make you AI native. That's like having a website and calling yourself a tech company lol. 5. No AI native org without AI native people. Most companies skip straight to the tools. That's why it fails. If your people don't understand how to manage agents, the tech doesn't matter. 6. Making your company "readable" to agents is the real work. Every process, every decision, every piece of knowledge needs to exist in a format an agent can consume. Most companies are nowhere close. 7. Speed without signal is just expensive chaos. You need the system to move fast AND know if you're moving in the right direction. 8. The skill chain is how agents get good at your specific workflows. Skills build on skills. The more you invest in them, the more your company compounds. 9. The moat is the system. People managing agents, agents reading from rich context, the whole thing getting smarter every week. That compounds. Your competitor can copy your tools. They can't copy your system. Full episode with Theo Tabah from LCA on The Startup Ideas Podcast (SIP) 🧃. This is the stuff we normally keep internal but all the sauce is yours. Theo Tabah is the brains behind advising the world's biggest companies on AI and building AI products. Your fav CEO's first call for figuring out AI. You are in for a treat Become AI native in under 60 minutes Watch

What does it actually mean to be AI native? There was no clear guide on the internet for how to become AI native so we built the definitive one (60 min masterclass): 1. An AI native org has 3 layers: people for strategy and taste, agents for execution, and a shared context layer that makes the entire company readable to agents. 2. AI eats the middle of your work. You used to spend 80% of your day on execution. Now agents do that. Your job is the bookends: deciding what to do and judging whether it's good enough. 3. Everyone is a manager now. Your output is the output of your agents. If your agents produce garbage, that's on you. You set them up wrong. 4. Using ChatGPT doesn't make you AI native. That's like having a website and calling yourself a tech company lol. 5. No AI native org without AI native people. Most companies skip straight to the tools. That's why it fails. If your people don't understand how to manage agents, the tech doesn't matter. 6. Making your company "readable" to agents is the real work. Every process, every decision, every piece of knowledge needs to exist in a format an agent can consume. Most companies are nowhere close. 7. Speed without signal is just expensive chaos. You need the system to move fast AND know if you're moving in the right direction. 8. The skill chain is how agents get good at your specific workflows. Skills build on skills. The more you invest in them, the more your company compounds. 9. The moat is the system. People managing agents, agents reading from rich context, the whole thing getting smarter every week. That compounds. Your competitor can copy your tools. They can't copy your system. Full episode with Theo Tabah from LCA on The Startup Ideas Podcast (SIP) 🧃. This is the stuff we normally keep internal but all the sauce is yours. Theo Tabah is the brains behind advising the world's biggest companies on AI and building AI products. Your fav CEO's first call for figuring out AI. You are in for a treat Become AI native in under 60 minutes Watch

GREG ISENBERG

85,083 views • 4 months ago

🚀 Exciting Launch on Product Hunt: Chat Thing! 🚀 Imagine creating a fully customized AI chatbot in minutes no coding required. That’s exactly what Chat Thing makes possible! Whether you want to train a bot on your own data or integrate with powerful AI models, this tool is designed to give you complete control over your chatbot experience. Why check it out? ✅ No-code setup – Build and deploy in minutes ✅ Custom AI training – Use your own data for personalized responses ✅ Seamless integrations – Connect with APIs, databases & more ✅ Versatile use cases – Customer support, personal assistants, and beyond If you've ever wanted to build your own AI assistant, now's the time. Check out Chat Thing on Product Hunt and let us know what you think! 🔗 Support Now: #AI #NoCode #Chatbots #ProductHunt #ArtificialIntelligence

🚀 Exciting Launch on Product Hunt: Chat Thing! 🚀 Imagine creating a fully customized AI chatbot in minutes no coding required. That’s exactly what Chat Thing makes possible! Whether you want to train a bot on your own data or integrate with powerful AI models, this tool is designed to give you complete control over your chatbot experience. Why check it out? ✅ No-code setup – Build and deploy in minutes ✅ Custom AI training – Use your own data for personalized responses ✅ Seamless integrations – Connect with APIs, databases & more ✅ Versatile use cases – Customer support, personal assistants, and beyond If you've ever wanted to build your own AI assistant, now's the time. Check out Chat Thing on Product Hunt and let us know what you think! 🔗 Support Now: #AI #NoCode #Chatbots #ProductHunt #ArtificialIntelligence

RAVI KUMAR SAHU

13,123 views • 1 year ago

Grok Voice Agent is absolutely insane Watch how it handles a dinner reservation flawlessly - felt exactly like talking to a real human host This is what AI-powered customer support looks like in 2026: → First response in under 1 second - 5× faster than rivals → 5 ultra-natural voices that talk, think, and act on your behalf → Full-duplex WebSocket for real-time, zero-lag conversations → Built-in live Web & 𝕏 search + MCP support → Auto-detects & switches between 20+ languages mid-conversation → 24/7 - never sleeps, never calls in sick And the base price is just $0.05/min For $3/hour, you get a 24/7 worker that never sleeps, never burns out, and treats every single customer with infinite patience Build voice agents for your business right now and replace your entire support queue Customer support is about to change forever xAI's voice agent is hands down the most impressive AI voice agent out there right now

Grok Voice Agent is absolutely insane Watch how it handles a dinner reservation flawlessly - felt exactly like talking to a real human host This is what AI-powered customer support looks like in 2026: → First response in under 1 second - 5× faster than rivals → 5 ultra-natural voices that talk, think, and act on your behalf → Full-duplex WebSocket for real-time, zero-lag conversations → Built-in live Web & 𝕏 search + MCP support → Auto-detects & switches between 20+ languages mid-conversation → 24/7 - never sleeps, never calls in sick And the base price is just $0.05/min For $3/hour, you get a 24/7 worker that never sleeps, never burns out, and treats every single customer with infinite patience Build voice agents for your business right now and replace your entire support queue Customer support is about to change forever xAI's voice agent is hands down the most impressive AI voice agent out there right now

X Freeze

591,858 views • 6 months ago

Most AI products added voice by wrapping a chatbot in speech-to-text. Vellum built a real-time voice loop on top of a full agent. It browses the web, reads your files, runs code, and sends messages while you’re still talking. This is voice as a real surface for an agent. Not a voice chatbot.

Most AI products added voice by wrapping a chatbot in speech-to-text. Vellum built a real-time voice loop on top of a full agent. It browses the web, reads your files, runs code, and sends messages while you’re still talking. This is voice as a real surface for an agent. Not a voice chatbot.

Kyronis

112,849 views • 3 months ago

Building real-time voice AI just got a lot easier 🤖✨ Introducing our new Python AI SDK for building real-time AI integrations directly into Stream Video. With out-of-the-box integrations with , Deepgram, , OpenAI, Moonshine AI, and more, developers can mix and match between real-time Voice AI, TTS or STT models. Get started 👇

Building real-time voice AI just got a lot easier 🤖✨ Introducing our new Python AI SDK for building real-time AI integrations directly into Stream Video. With out-of-the-box integrations with , Deepgram, , OpenAI, Moonshine AI, and more, developers can mix and match between real-time Voice AI, TTS or STT models. Get started 👇

Stream

224,795 views • 1 year ago

I've just tested the EVI 3 speech-to-speech model by Hume Hume AI, and this is the most realistic voice cloning I’ve heard. It doesn’t just clone the sound of your voice — it captures your tone, your speech patterns, your personality. Feels like your voice is thinking in real time. What’s wild is: — It’s not text-to-speech. It’s speech-to-speech. — You speak → it clones your voice + vibe — And it’s compatible with LLMs like Groq, Anthropic, Deepseek - Try talking to a voice clone: - Want more control? Customize it here: This changes how we’ll interact with AI voices. No more robotic replies — these sound human.

I've just tested the EVI 3 speech-to-speech model by Hume Hume AI, and this is the most realistic voice cloning I’ve heard. It doesn’t just clone the sound of your voice — it captures your tone, your speech patterns, your personality. Feels like your voice is thinking in real time. What’s wild is: — It’s not text-to-speech. It’s speech-to-speech. — You speak → it clones your voice + vibe — And it’s compatible with LLMs like Groq, Anthropic, Deepseek - Try talking to a voice clone: - Want more control? Customize it here: This changes how we’ll interact with AI voices. No more robotic replies — these sound human.

Aaliya

28,056 views • 1 year ago

Give your voice agents a face. Introducing Lemon Slice-2 – the world's first interactive talking AI video model Think of it as a face layer for your voice agents, available today in production as an API or an embeddable widget. What makes Lemon Slice-2 great: → Easy: All you need is 1 photo → Expressive: hands and full body motion → Any character and style: if it has a face, Lemon Slice-2 can animate it Under the hood: we trained a custom, 20B-parameter video diffusion transformer, streaming at 20fps on a single GPU. Infinite-length video generation with no error accumulation. Talk to the demo avatars on our homepage. Or, add your own interactive avatar to your website. Companies and creators are already building video agents that are: → Customer support agents who screen-share and show you where to click → Lead qualification agents that replace forms with real conversations → AI personal shoppers who try on clothes and guide purchases People don't engage with chat bubbles. They engage with faces. And now you can have a 24/7 spokesperson that’s the face of your brand. We’ve partnered with , Daily, and Modal to make this happen. As part of this launch, we’re giving away a guide: “How to Add a High-Converting Video Agent to Your Website in under 10 Minutes” Retweet and comment "LEMON SLICE" below and I'll send it over 🍋

Lina Colucci

1,163,590 views • 10 months ago

Watch me build + deploy a voice agent to the free tier of LiveKit Cloud in <2 mins. Latency is amazing, process is dead simple, and it's all your own code so your agent can do legit anything. This is the most flexible, performant, and EASY way to develop apps for voice.

Watch me build + deploy a voice agent to the free tier of LiveKit Cloud in <2 mins. Latency is amazing, process is dead simple, and it's all your own code so your agent can do legit anything. This is the most flexible, performant, and EASY way to develop apps for voice.

Shayne

11,335 views • 1 year ago

Launching today: the world’s first fully optimized LLM routing stack that can be embedded directly into your product, Merge Embedded Routing Stack. Model sovereignty is now table stakes for any AI product. Users and companies are demanding full control over the AI models running in your product. Right now, no company is able to do this without massive overhead. Until today. Merge makes it easy for your customers to use your product exactly the way they want to. With a single API, you can: 1. Allow customers to bring their own keys, manage their own budgets, and set policies like fallback rules, model allowlists, and more 2. Let each customer connect and switch their own models, right inside your product 3. Optimize each customer for cost, performance, or capability Without Merge, you’d be: - Building routing policies, provider integrations, and LLM key management - Giving your users zero model independence - Receiving endless customer support tickets asking “can I use 5.6 Sol?” Products that fight their customers' model strategy will lose to products that fit it. Merge is the only place that allows your users to produce the exact results they want, with the models they have access to. Give your customers the advantage, without the massive overhead costs.

Launching today: the world’s first fully optimized LLM routing stack that can be embedded directly into your product, Merge Embedded Routing Stack. Model sovereignty is now table stakes for any AI product. Users and companies are demanding full control over the AI models running in your product. Right now, no company is able to do this without massive overhead. Until today. Merge makes it easy for your customers to use your product exactly the way they want to. With a single API, you can: 1. Allow customers to bring their own keys, manage their own budgets, and set policies like fallback rules, model allowlists, and more 2. Let each customer connect and switch their own models, right inside your product 3. Optimize each customer for cost, performance, or capability Without Merge, you’d be: - Building routing policies, provider integrations, and LLM key management - Giving your users zero model independence - Receiving endless customer support tickets asking “can I use 5.6 Sol?” Products that fight their customers' model strategy will lose to products that fit it. Merge is the only place that allows your users to produce the exact results they want, with the models they have access to. Give your customers the advantage, without the massive overhead costs.

Shensi Ding

357,777 views • 3 months ago

Voice agents are awkward, and everyone notices: You ask a question. The agent thinks. You wait. And wait... Nobody wants this. I'd rather talk to a person. If your model's response time is over 300ms, you won't make it. Unfortunately, most text-to-speech models can't get anywhere close to that. I want you to take a look at the latest model released by : TTS-1.5. I built a simple voice agent using the model so you can see it in action and test it on your computer. You'll find the repository link below. The latency numbers of this model are wild: • Max model → under 250ms • Mini model → under 130ms That's 4x faster than prior generations and faster than human response times!

Voice agents are awkward, and everyone notices: You ask a question. The agent thinks. You wait. And wait... Nobody wants this. I'd rather talk to a person. If your model's response time is over 300ms, you won't make it. Unfortunately, most text-to-speech models can't get anywhere close to that. I want you to take a look at the latest model released by : TTS-1.5. I built a simple voice agent using the model so you can see it in action and test it on your computer. You'll find the repository link below. The latency numbers of this model are wild: • Max model → under 250ms • Mini model → under 130ms That's 4x faster than prior generations and faster than human response times!

Santiago

69,462 views • 9 months ago

EVERYONE IS BUILDING CUSTOM GPTS BUT ALMOST NOBODY IS TRAINING THEIR OWN MODEL FROM SCRATCH $100 and one weekend is all it takes with andrej karpathy’s nanochat you pretrain a gpt-2 grade base model on 400 billion tokens of web text, then fine-tune it on your own obsidian notes, your ai chat history, your writing it answers in your voice, recalls your own ideas, runs locally with no api and no subscriptions complete builder’s guide with every command in the article ↓

EVERYONE IS BUILDING CUSTOM GPTS BUT ALMOST NOBODY IS TRAINING THEIR OWN MODEL FROM SCRATCH $100 and one weekend is all it takes with andrej karpathy’s nanochat you pretrain a gpt-2 grade base model on 400 billion tokens of web text, then fine-tune it on your own obsidian notes, your ai chat history, your writing it answers in your voice, recalls your own ideas, runs locally with no api and no subscriptions complete builder’s guide with every command in the article ↓

leopardracer

15,859 views • 4 months ago

"Clippy is finally vindicated." Balaji Srinivasan: "You actually want to put faces on your AI avatars, on your AI agents." "The reason they're intuitive to a hundred million people is they're used to chatting with another human on the other side." "They're already modeling the chat box as being a human-like response." "But when it's outside of that chat box environment and it's suggesting on the screen, you kind of want a face to pop up." Balaji on The Network State Podcast

"Clippy is finally vindicated." Balaji Srinivasan: "You actually want to put faces on your AI avatars, on your AI agents." "The reason they're intuitive to a hundred million people is they're used to chatting with another human on the other side." "They're already modeling the chat box as being a human-like response." "But when it's outside of that chat box environment and it's suggesting on the screen, you kind of want a face to pop up." Balaji on The Network State Podcast

a16z

58,314 views • 8 months ago

xAI is cooking 🔥 They just launched Grok Voice, a platform for building AI voice agents. - You can build voice agents in under 2 minutes - No code required - Human-like conversations with sub-second latency - Supports 25+ languages - 80+ built-in voices or clone your own - Works with Gmail, Google Calendar, Outlook, Notion, MCPs, and more - Starts at $0.05/min and includes a free phone number

xAI is cooking 🔥 They just launched Grok Voice, a platform for building AI voice agents. - You can build voice agents in under 2 minutes - No code required - Human-like conversations with sub-second latency - Supports 25+ languages - 80+ built-in voices or clone your own - Works with Gmail, Google Calendar, Outlook, Notion, MCPs, and more - Starts at $0.05/min and includes a free phone number

AshutoshShrivastava

144,824 views • 4 months ago

NEW: Introducing Octane AI Agentic Commerce Quizzes - Increase sales with AI. What is it? A sales quiz AI agent that makes 1-1 personalized sales experiences for every single customer. In real time. Powered by our new AI model CORE-1. Examples: 📸 Want to ask your customer to take a selfie and your AI agent automatically recommends them a full outfit from your catalog? Octane AI agents can do that. 🪞 Want to have an AI agent hand pick out each product for a personalized skin care routine? Want them to upload a selfie to detect their skin tone? Octane AI agents can do can that. 📊 Want to create an incredibly detailed report with graphs and tables and graphics thats generated by AI for each customer? Octane AI agents can do that. We give you the building blocks and you can build anything. And you can build it fast because our AI will do the heavy lifting for you. This is v1 and a representation of where our commerce and quiz technology is headed. Available today to everyone at 🆕 What we are launching today: • Smart Quiz Builder: Have an AI agent plan out and build your Octane AI quiz for you. It can even write custom HTML for beautiful results pages and progress bars. • Smart Products: It can take forever to setup the recommendation logic for a quiz. For those of you who need help, simply add smart products to your Octane AI quiz and your very own AI agent will hand-pick products for each customer who takes your quiz. It’s amazing. • Smart Copy: Instead of showing everyone who takes your quiz the exact same copy, use AI to personalize the quiz for every single person who takes it. Explain why these specific products are perfect for specifically them. • Image Analyzer: Let your customers upload or take a photo during the quiz and have AI analyze it. You can use this for anything from skin tone detection to picking out outfits! • Shopping Assistant: An AI agent that lives on your store that can help your customers at the right time. We have been building quiz software for almost 10 years now and AI is enabling us to make quizzes even more powerful. This is just the v1 of what we will be releasing in this area. We are so excited to see what you create with these new agentic products. Get creative, we think you will be surprised at how many interesting experiences you can create with Octane AI now.

NEW: Introducing Octane AI Agentic Commerce Quizzes - Increase sales with AI. What is it? A sales quiz AI agent that makes 1-1 personalized sales experiences for every single customer. In real time. Powered by our new AI model CORE-1. Examples: 📸 Want to ask your customer to take a selfie and your AI agent automatically recommends them a full outfit from your catalog? Octane AI agents can do that. 🪞 Want to have an AI agent hand pick out each product for a personalized skin care routine? Want them to upload a selfie to detect their skin tone? Octane AI agents can do can that. 📊 Want to create an incredibly detailed report with graphs and tables and graphics thats generated by AI for each customer? Octane AI agents can do that. We give you the building blocks and you can build anything. And you can build it fast because our AI will do the heavy lifting for you. This is v1 and a representation of where our commerce and quiz technology is headed. Available today to everyone at 🆕 What we are launching today: • Smart Quiz Builder: Have an AI agent plan out and build your Octane AI quiz for you. It can even write custom HTML for beautiful results pages and progress bars. • Smart Products: It can take forever to setup the recommendation logic for a quiz. For those of you who need help, simply add smart products to your Octane AI quiz and your very own AI agent will hand-pick products for each customer who takes your quiz. It’s amazing. • Smart Copy: Instead of showing everyone who takes your quiz the exact same copy, use AI to personalize the quiz for every single person who takes it. Explain why these specific products are perfect for specifically them. • Image Analyzer: Let your customers upload or take a photo during the quiz and have AI analyze it. You can use this for anything from skin tone detection to picking out outfits! • Shopping Assistant: An AI agent that lives on your store that can help your customers at the right time. We have been building quiz software for almost 10 years now and AI is enabling us to make quizzes even more powerful. This is just the v1 of what we will be releasing in this area. We are so excited to see what you create with these new agentic products. Get creative, we think you will be surprised at how many interesting experiences you can create with Octane AI now.

Matt Schlicht

290,870 views • 11 months ago

Why most AI products fail: Lessons from 50+ AI deployments at OpenAI, Google & Amazon Aish and Kiriti Badam have built 50+ enterprise AI products across companies like OpenAI, Google, Amazon, and Databricks. Based on these experiences, they’ve developed a small set of best practices for building and scaling successful AI products. Our goal with this conversation is to save you and your team a lot of pain and suffering while building your AI products. We discuss: 🔸 Two key ways AI products differ from traditional software 🔸 Common patterns and anti-patterns when building successful AI products 🔸 Their framework for iteratively building AI products 🔸 Why evals aren’t a cure-all 🔸 Why obsessing about customer trust and reliability is an underrated driver of successful AI products 🔸 The skills that matter most for builders in the AI era Listen now 👇 • YouTube: • Spotify: • Apple: Thank you to our wonderful sponsors for supporting the podcast: 🏆 Merge — The fastest way to ship 220+ integrations: 🏆 Strella — The AI-powered customer research platform: 🏆 Brex — The banking solution for startups:

Why most AI products fail: Lessons from 50+ AI deployments at OpenAI, Google & Amazon Aish and Kiriti Badam have built 50+ enterprise AI products across companies like OpenAI, Google, Amazon, and Databricks. Based on these experiences, they’ve developed a small set of best practices for building and scaling successful AI products. Our goal with this conversation is to save you and your team a lot of pain and suffering while building your AI products. We discuss: 🔸 Two key ways AI products differ from traditional software 🔸 Common patterns and anti-patterns when building successful AI products 🔸 Their framework for iteratively building AI products 🔸 Why evals aren’t a cure-all 🔸 Why obsessing about customer trust and reliability is an underrated driver of successful AI products 🔸 The skills that matter most for builders in the AI era Listen now 👇 • YouTube: • Spotify: • Apple: Thank you to our wonderful sponsors for supporting the podcast: 🏆 Merge — The fastest way to ship 220+ integrations: 🏆 Strella — The AI-powered customer research platform: 🏆 Brex — The banking solution for startups:

Lenny Rachitsky

176,860 views • 9 months ago

Better/faster/cheaper voice AI turn detection with Gemini 2.0 The code that determines when the agent should respond to the user is some of the most important code in your voice AI agent. The technical terms for this job are "turn detection" or "phrase endpointing." If the voice AI responds before the user has finished their thought, the conversation is choppy and unproductive. If the AI waits too long, the conversation is slow and frustrating. There are a number of ways to approach this. You can: 1. Use a fast "voice activity detection" model to detect pauses in speech. Respond when the user pauses. 2. Use a specialized phrase endpointing model that operates on transcribed text, pattern-matching on text semantics. 3. Train a specialized phrase endpointing model that operates directly on audio. 4. Leverage the native audio capabilities of a SOTA LLM like Gemini 2.0. We've benchmarked all four of these, and Gemini 2.0 currently beats other approaches. Using Gemini is also cheaper than transcribing the audio separately using a transcription service or model. Here's a short video showing Gemini phrase endpointing in two scenarios. First, correctly handling pauses in natural conversation. Second, requesting a phone number (which is a common activity in a use case like customer support). You can see the Completeness check lines in the terminal output, printed each time Gemini processes a chunk of audio.

Better/faster/cheaper voice AI turn detection with Gemini 2.0 The code that determines when the agent should respond to the user is some of the most important code in your voice AI agent. The technical terms for this job are "turn detection" or "phrase endpointing." If the voice AI responds before the user has finished their thought, the conversation is choppy and unproductive. If the AI waits too long, the conversation is slow and frustrating. There are a number of ways to approach this. You can: 1. Use a fast "voice activity detection" model to detect pauses in speech. Respond when the user pauses. 2. Use a specialized phrase endpointing model that operates on transcribed text, pattern-matching on text semantics. 3. Train a specialized phrase endpointing model that operates directly on audio. 4. Leverage the native audio capabilities of a SOTA LLM like Gemini 2.0. We've benchmarked all four of these, and Gemini 2.0 currently beats other approaches. Using Gemini is also cheaper than transcribing the audio separately using a transcription service or model. Here's a short video showing Gemini phrase endpointing in two scenarios. First, correctly handling pauses in natural conversation. Second, requesting a phone number (which is a common activity in a use case like customer support). You can see the Completeness check lines in the terminal output, printed each time Gemini processes a chunk of audio.

kwindla

64,261 views • 1 year ago

Impressive use of #chatgpt to prepare a PowerPoint Presentation! Prompt: “I want you to write me VBA code for a PowerPoint presentation about the history of AI. You are to fill in all the text with your own knowledge, no placeholders. I need 5 slides.” Crt: Avin Foo #AI #tech

Impressive use of #chatgpt to prepare a PowerPoint Presentation! Prompt: “I want you to write me VBA code for a PowerPoint presentation about the history of AI. You are to fill in all the text with your own knowledge, no placeholders. I need 5 slides.” Crt: Avin Foo #AI #tech

Pascal Bornet

108,721 views • 3 years ago

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

kwindla

335,479 views • 2 months ago