Voice AI Meetup: Build Smarter Voice Agents with Context Carryover

​Universal-3.5 Pro Realtime is our newest streaming speech-to-text model, introducing context carryover—the first streaming STT model that uses the agent's previous question to better understand the user's response.

​Join the AssemblyAI team for a live walkthrough of the model, learn how the new capabilities improve real-world voice agents, and see how to tune accuracy, latency, and conversation quality.

​🔥👀 What's New with Universal-3.5 Pro Realtime

• ​Context carryover — agent_context feeds in the agent's question so the model hears the answer through it; 10.2% lower WER across a 20,000-file benchmark.

• ​Conversation memory, on by default — a rolling memory of the call with nothing to configure.

• ​Code-switching across 18 languages — mid-sentence switching, Hinglish included, plus language_code steering when you already know which one.

• ​Voice Focus & speaker labels — voice_focus isolates the primary speaker; labels re-cluster at stream end for up to 10 speakers.

​Agenda

• ​5:00–6:00 pm — Networking + Happy Hour

• ​6:00–6:15 pm — Live Demo 3.5 Pro Realtime

• ​6:15–7:00 pm — Fireside Chat + Q&A

• ​7:00-8:00 pm — Networking

​🔗 assemblyai.com | Check out 3.5 Pro Realtime →

​Limited Capacity

​Seats are limited. Request an invite to join the community building the future of voice AI!

​By signing up for this event, you opt in to receiving AssemblyAI marketing emails. You may unsubscribe at any time. By attending this event, you acknowledge that photography and video recording may take place; if you do not consent, please contact the event host.