Welcome (back) to AI Minds, a newsletter about the brainy and sometimes zany world of AI, brought to you by the Deepgram editorial team.

In this edition:

• 🎥 Ultimate AI Voice Cloning Guide - BEGINNER To Pro 2025

• 🔈 Voila: Voice-Language Foundation Models for Real-Time Autonomous Interaction

• 🧠 Analyzing Conversational Context Recall and Utilization in Voice Interaction Models

• 💻 Deepgram Saga: The Voice OS for Developers

• 🎯 What Developers Need to Know About Speech Recognition Metrics

• ⚡Deepgram Partners with AWS to Accelerate Voice AI Deployment

• 🐝 Social Media Buzz: Zero-Shot Voice Cloning, the smallest voice agent, and more!

• 💻 New Webinar: Building AI Voice Agents with Deepgram + AWS Bedrock

• 🎙️ The AI Minds Podcast with Thibault Mardinli, Explorer at Voice AI Space

• 📈 Gupshup raises $60M+ to expand its conversational AI and messaging platform

• 💰 Best AI Meeting Notes Assistants for Fintech Teams

• 🔊 In Case You Missed It: GPT-5 and the Future of Voice AI 

• 🤖 Voice Agent API Just Leveled Up: GPT-5 + GPT-OSS-20B

• 🍎 Apple’s new Siri may allow users to operate apps just using voice

• 💬 NVIDIA releases open dataset, models for multilingual speech AI

Thanks for letting us crash your inbox; let’s party. 🎉

🎥 Ultimate AI Voice Cloning Guide - BEGINNER To Pro 2025

Learn how to go easily from a beginner to a pro in this eleven-minute tutorial! The three tools featured are labeled as “Beginner AI Voice Cloning,” “Mid-Level AI Voice Cloning,” and “Professional AI Voice Cloning.” Which one do you think is best?

🔍 Voice-Language Foundation Models for Voice Role-Play and the Surprising Memory of Voice Assistants

Voila: Voice-Language Foundation Models for Real-Time Autonomous Interaction and Voice Role-Play - “The authors of this paper presentVoila, a family of large voice-language foundation models that moves beyond traditional pipeline systems by adopting a new end-to-end architecture that enables full-duplex, low-latency conversations while preserving rich vocal nuances such as tone, rhythm, and emotion.”

Does Your Voice Assistant Remember? Analyzing Conversational Context Recall and Utilization in Voice Interaction Models - “Recent advancements in multi-turn voice interaction models have improved user-model communication. However, while closed-source models effectively retain and recall past utterances, whether open-source models share this ability remains unexplored.” This paper fills this gap.

🧠 Deepgram Saga: The Voice OS for Developers

Deepgram Saga: Meeting You Where and How You Work - Execute complex workflows across your entire tech stack using voice or text—whatever feels natural in the moment. With one-click integrations to Gmail, Google Calendar, Slack, Notion, Linear, Perplexity, and more, Saga connects seamlessly to the tools you already use.

Deepgram Saga isn't just another voice assistant—it's a voice OS built specifically for developers. What you can do:

• 📅 Manage calendars and schedule meetings without leaving your IDE

• 📥Send emails, create tickets, and update team channels

• 🔍 Research solutions with Perplexity and save findings to Notion

• ⏩ The result? Faster execution, fewer context switches, and development that truly happens at the speed of thought.

Ready to make your voice your most powerful development tool?

Try Deepgram Saga →

⚡ What Developers Need to Know About Speech Recognition Metrics and Deepgram Partners with AWS to Accelerate Voice AI Deployment

What Developers Need to Know About WER, KER, and KRR - This article will introduce three of the most used and prominent speech recognition metrics: Word Error Rate (WER), Keyword Error Rate (KER), and Keyword Recognition Rate (KRR).

Deepgram Signs Strategic Collaboration Agreement with AWS to Accelerate Global Deployment of Voice AI - As a Generative AI Competency Partner and long-standing AWS Partner Network (APN) member, Deepgram offers a full-featured voice AI platform that includes speech-to-text (STT), text-to-speech (TTS), and speech-to-speech (STS) capabilities.

🐝 Social Media Buzz: Zero-Shot Voice Cloning and More!

Thibault Mardinli, Explorer at Voice AI Space. Voice AI Space is a beacon to master voice tech's wild seas. They guide developers, entrepreneurs, and enthusiasts to top tools, news, knowledge, and careers, empowering everyone in the vast voice AI ocean.

💻 New Webinar: Building AI Voice Agents with Deepgram + AWS Bedrock

Sign up here: ​Deepgram’s Voice Agent API brings lightning-fast speech-to-text and lifelike text-to-speech together with event hooks and speaker diarization, all in real time. Amazon Bedrock gives you instant access to leading foundation models like Claude and Titan, with built-in safety, compliance, and flexibility, perfect for powering voice agents with real intelligence.

​Join us to learn how to build scalable, responsive AI voice agents that actually work in production.

⏰ When: Tuesday, September 9, 10:30am - 11:30am PDT

🧭 Where: Online

✅ Sign up here!

🤖 Bonus Bits and Bytes!

• 🧠 Gupshup raises $60M+ to expand its conversational AI and messaging platform - Conversational AI continues to expand its presence, as we predicted in our State of Voice Report this year.

• 💰 Best AI Meeting Notes Assistants for Fintech Teams - Not all meeting notes assistants are made equal. Find out what the best ones are in this article.

• 🔊 In Case You Missed It: GPT-5 and the Future of Voice AI - Explore how GPT-5’s new capabilities redefine what is possible for real-time voice agents and the infrastructure needed to power them.

• 🤖 Voice Agent API Just Leveled Up: GPT-5 + GPT-OSS-20B - GPT-5 and GPT-OSS-20B are now live in the Deepgram Voice Agent API and Playground, giving developers more choice for reasoning depth, latency, cost efficiency, and open-source flexibility.

• 🍎 Apple’s new Siri may allow users to operate apps just using voice - Apple is testing a version of Siri that will be able to take actions on your behalf across various apps by following voice commands

• 💬 NVIDIA releases open dataset, models for multilingual speech AI- The new Granary dataset, featuring around 1 million hours of audio, was used to train high-accuracy and high-throughput AI models for audio transcription and translation