Corti's New Medical Speech Model Shatters Accuracy Records, Leaving OpenAI's Whisper in the Dust
Breaking: Corti's Symphony for Speech-to-Text Achieves 1.4% Word Error Rate on Medical Terminology
Copenhagen-based healthcare AI company Corti today launched Symphony for Speech-to-Text, a clinical-grade speech recognition model that achieves a record-breaking 1.4% word error rate (WER) on English medical terminology. This performance drastically outperforms generalist models, including OpenAI's Whisper (17.7% WER), ElevenLabs (18.1%), and Nvidia's Parakeet (18.9%).
The new model is designed for real-time dictation, conversational transcription, and batch audio processing in clinical settings. Corti claims its solution reduces errors by up to 93% compared to leading generalist APIs on medical terms, acronyms, and dosages.
Background
For decades, medical speech recognition focused on generating static text documents for physicians to review—a digital replacement for notepads. However, as healthcare enters the 'agentic era,' where autonomous AI agents assist in clinical decisions, the transcript becomes the foundational data layer for downstream reasoning.
General-purpose models like OpenAI's Whisper often fail on complex medical acronyms, shorthand, and noisy ER environments. Corti's Symphony aims to fill this gap by providing a specialized, production-grade API built from the ground up for clinical workflows.
Expert Quotes
“We are focused on ensuring our AI scribes can be trusted by physicians, medical practitioners, and patients—the entire healthcare system,” said Andreas Cleve, co-founder and CEO of Corti, in an exclusive interview. “Speech has always been one of healthcare’s most important inputs. What is changing is what happens after the words are captured.”
Cleve emphasized that in the agentic era, speech recognition must provide accurate clinical facts for AI reasoning. “If a model mishears a medication, dosage, or symptom, every downstream step becomes less reliable. Symphony for Speech-to-Text gives healthcare builders a speech layer accurate enough to thrive in clinical reality.”
What This Means
The launch signals a critical inflection point for healthcare developers. While general-purpose APIs are sufficient for broad-domain transcription, they introduce compounding risks when used as data inputs for clinical AI agents. Symphony's near-perfect accuracy on medical terminology reduces the chance of errors propagating through EHR navigation, decision support, and real-time patient monitoring.
This development underscores the value of domain-specific AI over foundation models in highly regulated industries. Corti's published research paper (see details) shows that specialized models can beat generalist providers when precision is paramount.
For hospitals and med-tech companies, adopting Symphony could mean fewer documentation errors, safer AI-driven recommendations, and greater trust from clinicians. The model is now available via API for integration into existing health IT systems.
Key Performance Comparison
- Corti Symphony: 1.4% WER
- OpenAI Whisper: 17.7% WER
- ElevenLabs: 18.1% WER
- Nvidia Parakeet: 18.9% WER
The data was published in a newly released research paper (link), demonstrating up to 93% reduction in word error rate on medical terminology compared to leading generalist APIs.
Related Articles
- Psychedelic Renaissance Skips Communities of Color, Experts Warn Despite Trump's Executive Order
- FDA Targets Weight Loss Drug Compounding: 8 Key Facts You Need to Know
- Tech's Toll on the Body: A Q&A with NPR's Manoush Zomorodi on 'Body Electric'
- Unlocking AI's Potential in Healthcare: Key Questions Answered
- 10 Game-Changing Insights from the Advanced Mediterranean Diet Study
- Isomorphic Labs: Alphabet’s $2 Billion AI Bet to Revolutionize Drug Discovery
- Hantavirus Outbreak on Cruise Ship Prompts Emergency Evacuation Plans in Canary Islands
- Debate Over Youth Social Media Bans Intensifies as Evidence Remains Unsettled