Cartesia September release notes

Hey [First name], More languages, more voices, more coffee. Ink-2 now supports five languages, the library grew by 85 voices, and we're hosting a full week of events (including coffee on us tomorrow) for SF Tech Week. Here's the rundown. What we shipped last month Ink-2 is now multilingual Ink-2 now transcribes French, Hindi, Japanese, and Spanish alongside English, with high accuracy and semantic endpointing in every language. And more on the way. Try it in the playground or through the API → 85 new public voices in the library Across 24 languages, including Hindi, Arabic, Indonesian, and Swedish. Check out the voice library → Greater control over pronunciation Pronunciation dictionaries on Sonic-3.6 now ignore case by default, so "new york," "New York," and "NEW YORK" share one entry. Need capitalization to change the meaning? Turn case sensitivity on for that entry. See custom pronunciations → See us at SF Tech Week Coffee CARTesia Wed, October 7, 9am–1pm PT: Order your free coffee with Sonic-3.6 and hang out with the team. RSVP → AI Builders Happy Hour Thurs, October 8, 6–8pm PT: Meet the team, connect with the community, and hear what we're building at our new HQ. RSVP → Voices in the Room with Cartesia and Coval Thurs, October 8, 6:30pm PT: A fireside chat on building and evaluating voice AI agents. Come for the talk, stay for the cocktails. RSVP → Vercel Ship Thurs, October 15, Palace of Fine Arts: Find us in the Startup Expo and see what your voice agent could look like in the future. Learn more → Drinks Around the Fire with Cartesia and Lorikeet Wed, October 28, 6:30–8:30pm PT: An evening of shop talk and lessons learned from teams building some of the most advanced voice agents. RSVP → Community & content Voice agents with Vapi: the replay Multilingual Agents Watch now → | Slides Early Intent Detection for Voice Agents Jev and Cartesia show how to detect what a caller wants before they finish speaking, so your agent can respond faster. Watch now → Build an AI Sales Coach A walkthrough of building a voice agent that helps B2B sales reps practice and sharpen their pitch. Watch now → From the blog Introducing Multilingual Voices: 50+ library voices now speak natively across 25 languages, so your brand sounds the same in every region. Why precision and recall matter so much: Why accuracy alone can mislead you when evaluating ML models, and how precision and recall give you an honest read on what your system gets wrong. TTS model vs. voice: they're not the same: The model generates the speech, and the voice shapes how it sounds. Knowing the difference helps you debug the right thing when your voice agent underperforms. Questions? Reply here or reach us at support@cartesia.ai You are receiving this email because you opted in to receive updates from Cartesia Cartesia, 555 De Haro St San Francisco, CA 94107 Unsubscribe