Mistral launches Voxtral, Krisp launches VIVA, Hume launches EVI3 and much more

Voice AI weekly digest

Top Updates 💪

• Mistral releases Voxtral, its first open source AI audio model (TechCrunch)

• Krisp launches VIVA SDK for improving turn-taking and achieves 1B mins/month milestone (SiliconAngle)

• Adobe's new AI tool turns silly noises into realistic audio effects (The Verge)

• Hume AI launches EVI 3, a speech-to-speech model that mimics voice, style (X)

• NVIDIA AI releases Canary‑Qwen 2.5B ASR‑LLM hybrid model (MarkTechPost)

• Nvidia released an open-source model Audio Flamingo 3 (MarkTechPost)

• Microsoft says this new voice conversion feature will improve AI dubbing (Slator)

• An open-source conversational AI platform, intervo.ai is now live (X)

• Zoho develops large language model with speech recognition (The Hindu)

• Enhancing multilingual speech with NVIDIA Riva TTS (NVIDIA Developer Blog)

• Vonage partners with AWS to unveil AI voice agent integration (PR Newswire)

• Deepgram receives 2025 voice AI excellence award (Morningstar)

• Squaretalk’s voice AI for fintech onboarding and fraud prevention (Fintech News)

• How synthetic voice is redefining multilingual communication (Telecom Reseller)

• AI voice tech boosts health insurance call center efficiency (Medium)

• UCaaS and CCaaS: A strategic move for the modern enterprise (UC Today)

• AI-powered TWS Earbuds: Real-time translation and transcription (Yanko Design)

• Peter Piper Pizza introduces voice AI phone ordering (SoundHound)

• Conversational banking: AI-powered financial CX (CX Network)

• SF AI startup Cluely wants to help you ‘cheat on everything’ (KRON4)

• ChatGPT Agent and the autonomous AI revolution (FourWeekMBA)

• Mistral AI adds voice AI tools to Le Chat (Artificial Intelligence News)

Voice AI Podcast 🎙️

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

---

Engineering Corner 😎

• Audio Flamingo 3 is now released (X)

• The pipecat-esp32 client runs on the AtomS3R (LinkedIn)

• ThinkSound by Alibaba for audio generation and editing (X)

• Open-source framework for real-time AI voice (X)

• Indic-Parler-TTS: 24 languages, 8000+ hours, 80k+ downloads (X)

• IndexTTS2: A Breakthrough in emotionally expressive and duration-controlled auto-regressive zero-shot TTS (Index‑TTS2)

• UniSLU: Unified spoken language understanding from heterogeneous cross-task datasets (arXiv)

• Fast inference end-to-end speech synthesis with style diffusion (MDPI)

• Got ChatGPT Plus? Record and summarize meetings on a Mac now (ZDNet)

• WhisperKit: On-device real-time ASR with billion-scale transformers (arXiv)

• Building voice agents: The revolutionary future of customer support is here (DEV)

• Top open‑source TTS models in 2025 (Modal)

• Speechmap.ai: Free speech dashboard for AI (SpeechMap.AI)

• Syllable.ai: Revolutionizing enterprise Voice AI automation (SaaStr)

• Whisper in the Wild: OpenAI's STT model in production (AI Mind)

• AI TTS could “unlearn” how to imitate certain people (MIT Technology Review)