Voice agents don’t sound natural when they interrupt too early or wait too long. Today we’re launching Flux, the first Conversational Speech Recognition model designed to solve this. Here’s the data that proves it 🧵👇

October: We launched Flux, the first conversational speech recognition model built to handle interruptions without trading off latency. We integrated the Voice Agent API with AWS Bedrock, partnered with teams like Lindy, and expanded Nova-3 again across Europe and Asia.

Full Self-Driving Supervised has many times more driving experience than any human driver It's trained on >100 years of data & collectively experiences a lifetime of driving scenarios every 10 minutes

July: We introduced Saga, our Voice OS for developers, designed to keep teams in flow using natural speech. Nova-3 earned the 2025 Voice AI Technology Excellence Award, and we expanded globally with Deepgram Dedicated and early access to our EU-hosted API. August: Nova-3

April: We crossed a major milestone with end-to-end speech-to-speech—no text conversion required—and introduced Aura-2, our most advanced, enterprise-grade text-to-speech model yet. May: Our partnership with Think41 showed what low-latency speech recognition enables in practice:

January: We entered 2025 cash-flow positive with 400+ enterprise customers, 3.3× usage growth over four years, 50,000+ years of audio processed, and more than one trillion words transcribed. February: We introduced Nova-3, setting a new standard for AI-driven speech-to-text

Are you tired of how much the US spends your money on military spending? See the full discussion via link in comments: @BruceFeinEsq @Dennis_Kucinich @Bitchute

🔊 This year, voice AI stopped being an experiment and became infrastructure. Looking back at our past 12 months, what stands out isn’t any single launch, but rather this: The month-by-month unfolding of our year👇

💰 Did this AI CEO's opinion on the "AI Bubble" change in the past two weeks? 📺 Feat: Scott Stephenson ⚡Title: "The Scott Stephenson AI Show | Episode 8" 🖇️ Watch here: https://t.co/peY1A9Hkx5

Is @Google's "Gemini 3 Deep Think" actually good? 📺 Feat: Scott Stephenson ⚡Title: "The Scott Stephenson AI Show | Episode 8" 🖇️ Watch here: https://t.co/peY1A9Hkx5

🤯 Why did @OpenAI recently pull the "Code Red" lever internally? 📺 Feat: @deepgramscott ⚡Title: "The Scott Stephenson AI Show | Episode 8" 🖇️ Watch here: https://t.co/peY1A9Hkx5

Humanoid robotics is no longer sci-fi; rising labor shortages and rapid AI advances are driving new interest in their development. The KOID ETF offers exposure to global companies in humanoid robotics and embodied AI innovation. Learn more about $KOID here.

🚀 How far does the latest Voice AI research push the boundaries of voice editing? Feat: @deepgramscott Title: "The Scott Stephenson AI Show | Episode 8" Full Episode: https://t.co/dCEe6OkTFa

🚀 Real-time Entity Detection is now live in Deepgram STT. Developers can now extract 50+ entity types in streaming transcription, including emails, names, phone numbers, dates, locations, SSNs, and more. Supported on Nova-3, Nova-2, Nova, and Enhanced models. Highlights: •

📈 Huge news! Deepgram just ranked among the Top 25 vendors by total customer spend in Brex's 2025 Benchmark report. The Brex Benchmark is a snapshot of the top software and AI vendors ranked by dollar spend from customers using Brex. We'd like to thank our friends at @brexHQ

📅 When: Wednesday, December 17th - 10 AM PT | 1 PM ET 💻 Where: Online 📝 Sign up here:

We've reached 1400+ registrants for this upcoming webinar! Last chance to sign up below 👇 A week ago, we only had 600 registrants. Thanks to you all, we've more than doubled that number. If you haven't signed up yet, you have less than 24 hours! ⏰

Here to Help® every step of the way.

📚 For more information, click here: https://t.co/OMiZQkBfar To try it out for yourself, check out the link to our Playground here: https://t.co/gi37PNtdxv

🌐 We thrilled to *finally* share this announcement... Aura-2 text-to-speech is now available in FIVE new languages! 🇳🇱 Dutch 🇫🇷 French 🇩🇪 German 🇮🇹 Italian 🇯🇵 Japanese These new languages are available in addition to English and Spanish. Text-to-speech is core to modern

📈 We just went from 600 registrations to over 1000 in less than a week! Don't miss this: Our upcoming live webinar is titled "Orchestrating Multi-Agent Voice AI Systems" Join this live, technical session to get a detailed breakdown of Deepgram’s Multi-Agent Framework, a

Link to playlist:

Flux Fridays — Week 2 If you’ve ever watched a voice agent miss its cue, talk over a user, or sit in awkward silence because the timing wasn’t right… this week’s songs are for you. This week’s theme: 🌀 Timing issues, latency moments, and the glitches they cause New

Catch a glimpse of Godzilla Survival, how to unlock Godzilla for FREE, and how to emotionally recover after losing your rings to Shockirus during the Godzilla Meets Sonic Rumble Crossover Event 🩹 Sonic Rumble’s Community Lead, Lukas, is back, and this time he’s explaining how

For more information, check out this blog: https://t.co/k6WTPGoz0G

🚀 We've been building up to this all year... 🌍 Earlier in 2025, we launched our EU-hosted API as part of Deepgram’s global expansion. Now, the Deepgram EU Endpoint is officially Generally Available. Thus, it's now even easier for teams to run their voice AI workloads fully