Events 🎤
AssemblyAI kicks off a hackathon week around its new TTS API (Oct 27 – Nov 7, Online, Luma)
Cartesia, Lorikeet, and Speak gather builders around the fire (Oct 28, SF, Luma)
Deepgram hosts Speak ‘26, a full day for people shipping voice AI (Oct 29, SF, Deepgram)
AssemblyAI launches its new text-to-speech model in a haunted house (Oct 29, SF, Luma)
SANE 2026 brings speech and audio researchers to MIT (Oct 30, Cambridge MA, SANE)
Top Updates 💪
Block unveiled Dessa, a full-duplex speech model trained only on audio that Jack Dorsey calls the most authentic yet. The Verge
ElevenLabs launched Eleven Realtime, a speech-to-speech model family that starts replying in about half a second. the-decoder
IBM released Granite Speech 5.0 as open-weight ASR under Apache 2.0, covering six languages with translation to English. MarkTechPost
Ant Group open-sourced Ming-3.0, five MIT-licensed speech models spanning TTS, audio understanding and recognition. Pandaily
Gradium open-sourced Phonon, a 31M-parameter on-device TTS model that starts speaking in about 7ms on a laptop CPU. MarkTechPost
Scaled Cognition launched Voice, a low-cost enterprise TTS model at $9 per million characters and 70ms latency. SiliconANGLE
Krisp is a finalist for Most Innovative CX Technology in CX Today’s 2026 CX Awards. LinkedIn
AWS upgraded the Polly generative engine, adding 14 voices plus Hindi and Finnish. AWS
OpenAI’s GPT-Live added prompt-to-voice, letting developers design a custom voice from a text description in one call. TechTimes
monday.com launched Harmony, a sales voice-agent platform using task-specific models at about 200ms latency. TNW
Synthflow went live on Zoom Phone and Contact Center, with Batteries Plus deploying it across 700+ US stores. Yahoo
Deutsche Telekom tapped AudioCodes to add AI call summaries, powered by Mistral Small 4, to business mobile plans. Telecompaper
Samsung added Malay to Galaxy AI, bringing Live Translate on calls to 24 languages. SoyaCincau
VoxNexus launched on-device voice AI for Android, claiming always-on listening at about 2.5% of battery over 12 hours. TradingView
Nirva raised $25M for an AI pendant that journals the wearer’s day and is built to capture only their voice. SiliconANGLE
ZDNet cloned a voice with free tools and ran a scam call on family, showing how accessible voice fraud has become. Fortune
Phonely made the case for measuring voice AI by revenue, citing customer bookings and conversions over cost savings. TMCnet
Voqalize argued voice is becoming the default software interface, pitching a voice layer that drops into existing apps. Business Standard
Engineering Corner 😎
AWS detailed streaming Kokoro TTS on SageMaker, synthesizing speech sentence-by-sentence as the agent generates it. AWS
A paper tested a frozen Gemma 4 speech encoder on Saudi and Gulf Arabic ASR, matching a larger model on character error rate. arXiv
A paper released OpenAMD, a 7-language answering-machine-detection dataset with a lightweight embeddings-plus-boosting detector. arXiv
Hlas-1 landed on Hugging Face, a 3.5B TTS model with zero-shot cloning that runs on a single 12GB GPU (non-commercial). Hugging Face
Wixzel shipped a Dart/Flutter voice SDK, for building agents that place and answer real phone calls over SIP. GitHub
A dev.to walkthrough added AI voice replies to Slack bots, wiring speech output into a bot workflow. dev.to