Speechmatics launches Medical Model for real-time clinical transcription

The potential of AI-powered medical transcription has always been transformative: reduce screen time, streamline documentation, and free clinicians to focus on patients instead of keyboards.

This promise is driving explosive growth as new players outcompete legacy incumbent technology while improving the doctor/patient experience – with the global market projected to reach $8.41 billion by 2032, and 30% of US healthcare providers expected to adopt ambient scribes by end-2025.

But most models crumble under the pressure of real clinical environments, mishearing drug names, mangling ICD codes, and struggling with the rapid-fire complexity of medical conversations.

Speechmatics' new Medical Model is built to change this. 

As the leader in the real-time streaming speech-to-text (STT) space, Speechmatics has released this domain-specific model for HealthTech partners requiring advanced accuracy, particularly across broad medical terminology and in clinical environments where live note-taking and dictation are critical.

Already outperforming the competition

Speechmatics has consistently outperformed specialist medical model providers across a broad range of languages in both file-based and streaming modes.

The new Medical Model delivers 98% accuracy - that's twice as good as your next best option - positioning Speechmatics further ahead of other live streaming STT providers.

Why this model matters now

In healthcare, conversations move fast.

They're full of shorthand, acronyms, multiple speakers and critical decisions made in seconds. 

Most transcription tools are trained for structured dictation, but real clinical dialogue is something else entirely – unstructured, rapid-fire, and packed with context that changes meaning.

This challenge is compounded by the accelerating pace of medical knowledge: what took 50 years to double in 1950 now doubles every 73 days, making AI-driven support systems critical to help clinicians process and apply this overwhelming volume of information in real-time conversations.

Speechmatics’ Medical Model is tuned for these real-world conversations, where speed, context, and precision are fundamental.

It's designed to support use cases where transcription needs to keep pace with care, such as:

• Contact center triage

• Ambient scribes and EHR integrations

• In-room consultations and dictation

• Telemedicine appointments

• Radiology voice notes

The Medical Model pushes transcription accuracy to new levels in the moments that matter most. Available now in English, multilingual support across Speechmatics' 55+ supported languages is coming soon.

Tuned to understand medical context

In reality, medical professionals and patients speak in a range of accents. Your model needs to be able to understand every voice. This is not the case for the majority of models, which struggle with anything beyond standard pronunciation and mainstream voices.

Speechmatics' Medical Model, however, excels where others falter. Built on our accent-independent foundation that recognizes voices across 55+ languages, the model handles the linguistic diversity of modern healthcare – from emergency calls with heavy regional accents to consultations with international specialists. 

Our system delivers clinical-grade accuracy regardless of how doctors, nurses, or patients actually speak.

This foundation enables our new model to recognize and correctly transcribe:

• ICD-10-CM codes

• Clinical acronyms and shorthand

• Drug names and procedures

• Prescriptions and dosages

• Phone numbers, addresses, and patient names

Essential to accurate clinical documentation, these areas are often misrecognized by both specialized and general-purpose models.

The Medical Model delivers superior accuracy for live workflows compared to traditional batch processing approaches.

Clinical impact: where performance meets usability

Real-world data reveals the scale of impact. A recent JAMA study found virtual scribes cut total EHR time by 5.6 minutes per appointment, including 1.3 minutes less note-taking and 1.1 minutes less "pajama time" (after-hours documentation). 

Research has also shown that "power users" see even bigger gains: a follow-up evaluation showed the top 33% of clinicians (by usage) achieving 2.5× greater time savings per note than lower-frequency users.

This proven reliability extends to mission-critical settings like ambulance services, where Speechmatics currently powers transcription for every emergency call across the UK, operating under a 100% uptime guarantee where reliability is literally a matter of life and death.

For clinical teams and healthtech providers, this translates to faster, cleaner EHR integration, less time spent correcting errors, and reduced friction in patient interactions.

"We're seeing great results already with Speechmatics. It's performing well in live deployments and accelerating our clients' efficiency at scale."  – Martin Taylor, CTO, Content Guru

Deployment, security, and availability

Healthcare data demands the highest security standards, and the Medical Model delivers comprehensive protection:

• HIPAA-compliant infrastructure

• Zero data retention

• Encryption in transit and at rest

• Audit trail support

• Available now in preview via SaaS

• On-prem availability for enterprise customers

The model is currently available in English, with rollout across Speechmatics' 55+ supported languages planned next.

Partners working in regulated healthcare environments have found confidence in Speechmatics' approach to compliance.

As a global ambient scribe customer shared "being in a regulated industry means data security, privacy, and compliance are top priorities for us. We've been reassured by how Speechmatics handles sensitive information and follows strict standards. The willingness of the team to understand and accommodate these regulations has really stood out and given us the peace of mind we need to stay compliant in all our processes."

Now accepting preview partners

If you're developing healthcare applications – from dictation tools and scribe systems to AI assistants, and need a transcription engine optimized for real-time medical speech, the Speechmatics Medical Model is now available to test.

Ready to see clinical-grade transcription in action? Head to Speechmatics' Portal to use the Medical Model directly in preview.