Streaming STT Performance Update | August 1, 2025 Newsletter
AssemblyAI's new speaker diarization model delivers 30% better accuracy in noisy audio. Plus: Build 465ms voice agents & learn how Dovetail improved WER by 36%.



Streaming Speech-to-Text Improvements: What's New
AssemblyAI has released significant performance improvements for our Streaming Speech-to-Text API, delivering substantial error rate reductions in critical transcription areas. These updates are now live for all users.
Key Performance Improvements for Streaming STT
Our latest streaming improvements have achieved significant performance gains on clean data, with a focus on accuracy for repeated digits and tokens.
Repeating Digits and Tokens Enhancement
What We Fixed:
- Missing repeated digits in transcription output
- Repetitive token recognition issues (e.g., "yes" "yes")
Performance Metrics:
- Previous error rate: 28.20%
- New error rate: 13.47%
- Total improvement: 52% reduction in error rate
Real-World Impact: This enhancement provides better handling of:
- Phone numbers
- Confirmation codes
- Account numbers
- Repetitive speech patterns
Latest AssemblyAI Resources and Tutorials
Real-Time Conversation Intelligence Guide
Discover how real-time conversation intelligence is transforming customer interactions from post-call analysis to live insights. Our comprehensive guide explores how streaming speech-to-text enables proactive engagement and immediate issue resolution.
Read the full guide: Real-time conversation intelligence
Hotword Detection Tutorial with Go and Streaming Speech-to-Text
Learn how to implement hotword detection using AssemblyAI's Universal-Streaming Speech-to-Text API and Go programming language. This tutorial covers:
- WebSocket integration for real-time streaming
- Real-time voice recognition implementation
- Custom hotword trigger configuration
- Practical applications for voice-activated systems
Access the tutorial: Hotword detection with AssemblyAI streaming speech-to-text
Video Tutorial: Building an AI Meeting Scheduling Assistant
Our latest YouTube tutorial demonstrates how to build an AI-powered meeting scheduling assistant using AssemblyAI's Speech-to-Text API. The step-by-step guide includes:
- Voice input integration with AssemblyAI
- Natural language meeting request parsing
- Automated calendar availability checking
- Programmatic meeting invitation sending
Watch on YouTube: Build an AI Assistant for Meeting Scheduling
Try AssemblyAI's Improved Streaming Performance
Experience the enhanced streaming accuracy firsthand in the AssemblyAI Playground. Upload your audio files or test with our examples—no coding required.
Get Started with AssemblyAI Streaming Speech-to-Text
Ready to implement these improvements in your application? Here are your next steps:
- Review the Documentation: Visit our Streaming Speech-to-Text API docs
- Test in the Playground: Try the AssemblyAI Playground with your own audio
- Join the Community: Connect with developers on Discord
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.