Making voice AI conversations actually conversational (with STT, TTS, agents, models, and all that)
San Francisco, CA
Joined February 2016
- Tweets2.9K
- Following745
- Followers11.6K
- Likes2.4K
Pinned Tweet
Building better voice AI should also mean giving developers more choices.
Today at #DeepgramSpeak, @trydaily, the company behind the open source @pipecat_AI framework, is joining Deepgram. We're thrilled to welcome @kwindla, our new Head of Open Source, and his team to Deepgram.
We’ve also co-developed the Real-Time Context Standard with @pipecat_AI, developed to help voice AI systems carry the context that makes conversations work.
More on the vision we’re working toward together: deepgram.com/blog/daily-pipe…
Wanna talk talk? We're heading to @vapi_ai’s VapiCon, Nov. 11-12 in SF. Catch our team on stage: CEO @deepgramscott, CPO @edanuff, and our new Head of Open Source @kwindla will all be in the room. See you there: vapicon.ai/
We're excited to have Scott Stephenson, Founder & CEO of Deepgram , returning to the stage for a Keynote presentation at VapiCon 2026!
Scott is a dark matter physicist turned Deep Learning entrepreneur. He earned a PhD in particle physics from the University of Michigan where his research involved building a lab two miles underground to detect dark matter. Scott left his physics post-doc research position to found Deepgram.
We can't wait to see you in San Francisco this November to hear Scott and other industry leaders discuss the future of Voice AI.
📍 Fort Mason, San Francisco
📅 November 11–12, 2026
🎟️ Get your ticket: vapi.ai/vapicon-2026
More choice, less searching.
Deepgram Marketplace is now in early access: one place to find the tech behind your voice agents without starting a new vendor hunt each time.
dpgr.am/mty402fu
Last week, we launched two LLMs at #DeepgramSpeak:
→ Think-Live LLM, which manages the conversation in real time and decides when to listen, wait, or respond
→ Think-Fast LLM, which does the work behind the conversation: following your rules, making tool calls, and writing replies built to be spoken
Add them to STT, TTS, and our Real-time Context Standard (RCS) for richer context, and you've got yourself a multisensory, real-time voice agent that understands every um, er, pause, & interruption. See how in this demo from Daniel Mishler, Head of Deepgram Labs. And learn more from our Chief Product Officer, @edanuff, on the blog: deepgram.com/blog/voice-agen…
Same Deepgram. Only louder.
Speaking is the most natural way to communicate. But for years, technology has asked us to meet it halfway.
Type this. Pick from a menu. Say it again, slower.
Deepgram has spent years closing that gap. Now our brand shows it.
Voice AI needed a brand worth talking about. Well, we’re here.
Let’s talk.
Building a real-time voice agent that holds a natural conversation requires a lot of orchestration under the hood.
Join our VP of Engineering Kris Efland and @pvergadia on Nov. 5 for a live look at the architecture behind production-ready voice AI.
Register: maven.com/p/eaefbf/build-rea…
We’re thrilled to welcome the @trydaily team and our new Head of Open Source @kwindla to Deepgram.
We’re bringing together our voice AI infrastructure and models with their open source and vendor-neutral framework, @pipecat_AI, to create an ecosystem where everyone can build voice agents using the best tools available.
Thank you @IndianIdle and @TechCrunch for helping us share the news.
techcrunch.com/2026/10/29/de…
That’s a wrap on #DeepgramSpeak. But we’re just getting started. Thanks to everyone who came out to talk, plan, and build the future of voice with us.
12
666
Our first batch of keynote speakers at #DeepgramSpeak took us through the journey of a voice AI agent.
@ToastTab showed us how to ship and scale for the enterprise.
@cresta talked about putting agent adaptability into practice using behind-the-scenes context.
And @doximity, @CoxAutomotive, and @UiPath envisioned a world where voice AI is the core way people interact with a product.
We’ve learned a lot, and the day’s not over yet.
Meet Think-Live LLM and Think-Fast LLM, two new LLMs built for voice agents.
Think-Live runs the conversation: when to listen, wait, or respond.
Think-Fast does the work: rules, tools, spoken replies.
Our CPO @edanuff breaks down this, and more from #DeepgramSpeak, in his latest blog: deepgram.com/blog/voice-agen…
"In voice AI, real-time context is most important."
For voice agents to hold a real dialogue, the models handling listening, thinking, and speaking need access to the same real-time context.
At #DeepgramSpeak, @deepgramscott introduced the Real-Time Context Standard: an open source approach to building agents that combine control and expressivity for authentic conversations.
More to come — stay tuned.
#DeepgramSpeak starts today. We're bringing together the people building the new wave of voice AI.
Stay tuned.
Real talk is the new bar. Tomorrow is #DeepgramSpeak, where we’re bringing together the people building voice AI to explore what's possible when machines keep up with real conversation. We’ll see you there.
Top-notch voice AI keeps the conversation flowing while getting things done behind the scenes. However, making it all feel effortless takes serious coordination. We're digging into what it takes to make human + AI interactions feel natural at #DeepgramSpeak this Thursday in SF. Join us and get $500 in Deepgram credits: deepgram.com/speak#registrat…
Your user pauses to think. Your agent jumps in.
Why is context so vital for voice AI to get the full picture?
@kwindla of @trydaily and @pipecat_ai gets into it at #DeepgramSpeak.
Attend and get $500 in Deepgram credits: deepgram.com/speak#registrat…
Transcript’s clean. Agent still got it wrong.
Some of the worst voice AI bugs hide in the handoffs between components, where one small disconnect derails everything.
6 days to #DeepgramSpeak, Oct. 29 in SF. Attendees get $500 in Deepgram credits: deepgram.com/speak#registrat…
3
9
679
Models aren’t what’s holding voice AI back, says @jordan_dearsley. Distribution is.
The @Vapi_AI CEO joins @deepgramscott to dig into that, plus what devs are still struggling to build with voice.
#DeepgramSpeak, Thursday, Oct. 29, 4:30 PM in SF: deepgram.com/speak#registrat…
“That’s fine.”
Agreement? Frustration? Text alone can't tell you.
Tone, pauses, and history all change what users mean. At #DeepgramSpeak (Oct. 29, SF), we’re digging into how voice agents can keep up.
Attend and get $500 in Deepgram credits: deepgram.com/speak#registrat…
2
7
551
Want a peek of what’s to come at #DeepgramSpeak? @covaldev’s @bnicholehopkins and @kwindla of @trydaily caught up ahead of the big day to chat about what it’ll take to build better voice agents.
Join them in SF on Oct. 29 and get $500 in Deepgram credits: deepgram.com/speak#register
Deepgram is hosting a voice agents developer conference in San Francisco on October 29th. I'm speaking at the conference, we have some fun Pipecat announcements to make, and lots of friends are either attending, or speaking, or both.
I figured this was an opportunity to record some conversations with people in the voice AI ecosystem about what they're interested in right now, what they're talking about at Speak, and what's new since we last caught up.
@bnicholehopkins, founder of @covaldev, has been building simulation and testing tools for voice agents since 2024. Before Coval, Brooke worked on simulation tooling at Waymo.
Brooke and I riffed about how deployments of enterprise voice AI are growing incredibly fast right now, how we think about building with open source components and models, applying systems engineering principles to agentic AI development, and aiming for "self-improving agent loops."
I've been experimenting with prompting Opus to create motion graphics and video effects. A discerning 11-year old of my acquaintance called the result, here, "kind of cringe." But I kind of like it. See what you think. :-)