Sam Altman is right and very wrong about AI-faked voices. And much more!

Voice AI weekly digest

Top Updates 💪

• OpenAI CEO Sam Altman is right and very wrong about AI-faked voices (WashingtonPost)

• Microsoft is doubling down on multilingual large language models – and Europe stands to benefit the most (ITPro)

• Voice AI landscape in Europe 2025 (Sifted)

• Freed says 20,000 clinicians are using its medical AI transcription ‘scribe,’ but competition is rising fast (VentureBeat)

• Amazon is acquiring Bee, maker of a wearable AI assistant that listens to conversations (Geekwire)

• Voice Fraud Prevention in the Age of AI and Hybrid Work (UCToday)

• Gupshup raises $60M+ to expand its conversational AI and messaging platform (SiliconAngle)

• AI voice company Hyper raises $6.3M to help automate 911 calls (TechCrunch)

• Voxtral technical report (X)

• Why AI Should Prioritize Conversations Over Automation in Outbound Sales (UniteAI)

• AI Voice Assistants in UC: Build, Buy, or Bridge? (UCToday)

• Speechmatics shipped realtime speaker diarization for voice agents (X)

• How AI speech-to-text technology is tuning in to a digital Saudi Arabia (ArabNews)

• Best AI Meeting Notes Assistants for Fintech Teams (Medium)

• SayWrite.ai Is Redefining Productivity with AI-Powered Voice Note Taking (Medium)

• Amplify Launches Custom AI-Powered Automatic Speech Recognition System (TheJournal)

• Hume AI delivers speech models on SambaCloud (SambaNova)

• Lightning Captions: Real-Time Transcription and Translation for the Classroom (PRNews)

• Leena AI unveils conversational AI ‘colleagues’ for the enterprise (ComputerWorld)

Voice AI Podcast 🎙️

In case you missed the latest episode of Voice AI Podcast…

---

Engineering Corner 😎

• Introducing Version 2 of Higgs Audio Generation (Boson)

• NAR-SREC: Nonautoregressive End-to-End Speech Recognition With Error Correction Decoder

• RapFlow-TTS: Rapid and High-Fidelity Text-to-Speech with Improved Consistency Flow Matching

• ZipVoice: Fast and High-Quality Zero-Shot Text-to-Speech with Flow Matching

• Conan: A Chunkwise Online Network for Zero-Shot Adaptive Voice Conversion

• Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models

• Mureka TTS V1: A new voice model from Mureka AI, integrated into their AI music and audio platform

• Micdrop v2 Launch: Micdrop, an open-source set of TypeScript packages for building real-time voice conversations with AI agents, launched its v2

• macos-local-voice-agents: A new open-source GitHub repository was introduced for running Pipecat voice AI agents locally on macOS

• Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance

• Speaker Disentanglement of Speech Pre-trained Model Based on Interpretability

• SpecASR: Accelerating LLM-based Automatic Speech Recognition via Speculative Decoding.