GoTranscript
>
All Services
>

Public/assemblyai Ships Universal 3 Pro Voice Agent Api

AssemblyAI ships Universal-3 Pro, Voice Agent API (Full Transcript)

New streaming transcription model, medical mode, real-time voice agent WebSocket API, LLM Gateway additions, language upgrades, and enhanced PII redaction.
Download Transcript (DOCX)
Speakers
add Add new speaker

[00:00:00] Speaker 1: Hi everyone, it's been a crazy couple of months here at Assembly and I figured I'd make a short update video for you to get you caught up to speed on everything that we've released in the last couple of months. First up is Universal 3 Pro Streaming. Universal 3 Pro Streaming is our latest and greatest streaming model with amazing accuracy, especially on key entities like emails, phone numbers and proper nouns. Check this out. My email address is martinschweiger at assemblyai.com. My phone number is 91009200. My credit card number is 4001-5012-5013-5014 and my CVV is 420. And it gets even better if you're in the medical space because we just released medical mode for even better performance on medical terminology without the need to find and configure your own key terms. With one parameter you get the best performance for medical scribes and note takers. This next launch was my favorite. Our voice agent API is finally here. It is the easiest way to build high quality voice agents with a simple WebSocket API where you send us audio and we send you audio back. We handle everything it takes to build a great voice agent under the hood like tool calling, turn detection, reconnection and so much more. If you haven't already, I really recommend that you give our voice agent API a try. It's really easy to get started and build something great with our API especially now that we've released our DOCKS MCP server and our Cloud Code skill. The full instructions for installation in your development environment is in our DOCKS so check it out to ensure that your agents are updated with the latest information about Assembly AI. Next we released a bunch of new models on LLM Gateway including Opus 4.7, Kimi 2.5, Quen 3 and GVT 5.5. We shipped a bunch of improvements to Universal 2 for Hebrew and Swedish and we have improvements coming for more languages very soon. And we shipped new parameters for PII redaction that let you retrieve both a redacted and unredacted text in one transcript request. As always the best way to keep a pulse on all of our launches is on our changelog. But that was all I had for this video. I hope you found that helpful and I'll see you in the next one. Bye.

ai AI Insights
Arow Summary
AssemblyAI shares a product update covering recent launches: Universal-3 Pro Streaming with improved accuracy on entities (emails, phone numbers, proper nouns) plus a new medical mode for better medical terminology transcription; the new Voice Agent API via WebSocket that supports audio in/out and handles tool calling, turn detection, reconnection, and more; tooling updates including a Docs MCP server and Cloud Code skill to keep agents current; new LLM Gateway models (Opus 4.7, Kimi 2.5, Qwen 3, GPT 5.5); Universal-2 improvements for Hebrew and Swedish with more languages coming; and new PII redaction parameters allowing both redacted and unredacted text in a single transcript request, with changelog recommended for ongoing updates.
Arow Title
AssemblyAI Product Update: Universal-3 Pro, Voice Agents, and More
Arow Keywords
AssemblyAI Remove
Universal-3 Pro Streaming Remove
medical mode Remove
speech-to-text Remove
transcription Remove
entity recognition Remove
Voice Agent API Remove
WebSocket Remove
tool calling Remove
turn detection Remove
LLM Gateway Remove
Opus 4.7 Remove
Kimi 2.5 Remove
Qwen 3 Remove
GPT 5.5 Remove
Universal-2 Remove
Hebrew Remove
Swedish Remove
PII redaction Remove
changelog Remove
Docs MCP server Remove
Cloud Code skill Remove
Arow Key Takeaways
  • Universal-3 Pro Streaming improves accuracy, especially for key entities like emails, phone numbers, and proper nouns.
  • A new medical mode boosts medical terminology performance without custom key-term configuration.
  • AssemblyAI launched a Voice Agent API that provides a simple WebSocket interface for real-time audio in/out voice agents.
  • The Voice Agent API abstracts key agent infrastructure such as tool calling, turn detection, and reconnection.
  • New enablement tooling includes a Docs MCP server and a Cloud Code skill to keep agents updated with AssemblyAI docs.
  • LLM Gateway added new models: Opus 4.7, Kimi 2.5, Qwen 3, and GPT 5.5.
  • Universal-2 received language improvements for Hebrew and Swedish, with more languages planned.
  • New PII redaction options allow retrieving both redacted and unredacted transcripts in a single request.
  • The changelog is the recommended place to track ongoing releases.
Arow Sentiments
Positive: Upbeat, promotional tone focused on new launches and improvements; emphasizes 'latest and greatest,' 'favorite,' and ease of getting started, with excitement about performance gains and new capabilities.
Arow Enter your query
{{ secondsToHumanTime(time) }}
Back
Forward
{{ Math.round(speed * 100) / 100 }}x
{{ secondsToHumanTime(duration) }}
close
New speaker
Add speaker
close
Edit speaker
Save changes
close
Share Transcript