Oliver Parish

  Machine Learning Engineer

A technical deep-dive into Token Duration Transducers (TDT) — the frame-skipping architecture behind Nvidia's Parakeet models. Covers inference mechanics, training with forward-backward algorithm, and how TDT achieves up to 2.82x faster decoding than standard RNN-T.

new blog image header

How Nvidia Dominates the HuggingFace Leaderboards in This Key Metric

Token Duration Transducer (TDT) Explained: How Frame-Skipping Achieves 2.8x Faster ASR

Technical

Edvak author

Founder and CEO, Edvak EHR

As clinical workflows become automated and AI-driven, real-time speech is shifting from a transcription feature to the foundational intelligence layer inside modern EHR systems.

[alt: Healthcare professionals in scrubs and lab coats walk through a hospital corridor. Some are holding documents or tablets, while a patient on a stretcher is wheeled past. The setting is busy, suggesting a fast-paced medical environment.]

Edvak case study

Why AI-native EHR platforms will treat speech as core infrastructure in 2026

Use Cases

Speechmatics Logo New

Editorial Team

Turning real-time clinical speech into trusted, EHR-native automation.

[alt: Logos of Speechmatics and Edvak are displayed side by side, interconnected by a stylized x symbol. The background features soft, wavy lines in light blue, creating a modern and tech-focused aesthetic.]

Speechmatics x Edvak - Header (1200 × 900)

One word changes everything: Speechmatics and Edvak EHR partner to make voice AI safe for clinical automation at scale

Company

Two European AI leaders combine forces to deliver responsible, enterprise-grade technology for financial services, healthcare, and public sector.

boost.ai partners with Speechmatics

Speechmatics and Boost.ai partner to power enterprise Voice AI for Europe's most regulated industries

News

Tom Young

Digital Specialist

Voice AI moved fast. Regulation caught up. Here’s how to stay compliant in 2026.

[alt: A computer server sits at the center, surrounded by two monitors displaying a padlock and a user icon. A figure works on a laptop, while a smartphone and security icons are nearby, emphasizing themes of cybersecurity and user interaction.]

compliance-voice-ai - Header (1200 × 900)

Your essential 2026 guide to voice ai compliance in today's digital landscape

Maria Anastasiou

Events & Customer Marketing Lead

What teams deploying Voice AI at scale see coming next as production pressure reshapes adoption.

Predictions for Voice AI in 2026

7 Voice AI predictions from teams building at scale in 2026

Yahia Abaza

Senior Product Manager

Expanding a Nordic medical lineup with 3.91% KWER model that delivers sub-second latency across Swedish, Finnish, Danish, and Norwegian clinical workflows.

[alt: Two healthcare professionals, wearing blue scrubs, engage in conversation in a hospital in Sweden]

Swedish medical model roll out - Header (1200 × 900)

Speechmatics launches new Swedish medical model, cutting transcription errors by 40%

Product

The numbers from 2025 show voice AI has moved from demos to dependable infrastructure, where ROI, trust, and real-time performance decide who wins in 2026.

Voice AI in 2026

Voice AI in 2026: 9 numbers that signal what's next

General

Archie McMullan profile picture

Speechmatics Graduate

Introducing 250ms final transcripts for Voice AI

Blog header image FEOU

You can’t hurry love, but you can hurry final transcripts

Built on NVIDIA AI infrastructure, the collaboration combines medical-grade speech models and agent workflows to deliver measurable ROI; cutting admin time and scaling healthcare capacity worldwide.

Header-Sully-announcement

Speechmatics and Sully.ai partner to scale healthcare AI infrastructure globally

Oak is Speechmatics’ new medical speech-to-text model, built for clinical conversations that move between languages and speakers.

Meet Oak: medical transcription that follows the conversation across languages and speakers

Sign In

Get started

Contact sales

Speech-to-Text API

Capabilities

Text-to-Speech API

Agent STT

Speech-to-text

Medical and healthcare-Navigation icon

Medical & Healthcare

Legal icon

Legal Transcription

technical-blog-icon

AI Voice Agents

media-and-captioning-icon

Media Distribution & Captioning

meeting-platforms

Note-Taking & Meeting Assistants

contact center icon

Contact Center Solutions

By Use Case

blog-icon

News & Articles

events-icon

Partnerships

contact-us-icon

Contact Us

careers-icon

Careers

about-speechmatics-icon

Find Out More About Speechmatics

case-studies-icon

Case Studies

Resources

Technical information

documentation-icon

service-status-icon

Service Status

Inclusivity

Community

Technical blog

announcement-icon

GitHub Academy

market-exposure-icon

Startup Program

Developers

Enterprise

Security

Pricing

Advanced speech recognition API with high accuracy across [<langcount>] languages.

Speechmatics Speech-to-Text

Get started with our speech recognition API

Free

Speechmatics

Advanced features for professional use ($0.24 per hour)

Custom solutions for large-scale deployments

We are Speechmatics. Global experts in deep learning and voice technology. Accurate speech-to-text for analysis and accessibility.

Speechmatics Social Sharing

Global experts in deep learning and voice technology. Accurate speech-to-text for analysis and accessibility.

Stay up-to-date with the very latest product news, company news and artificial intelligence (AI) research. Bookmark our technology blog today!

Blog & Latest Speech Recognition News| Speechmatics

Discover how medical transcription improves clinician productivity by reducing documentation time, streamlining workflows, and supporting patient care.

How Medical Transcription Supports Clinician Productivity

As healthcare systems face growing pressure, multilingual AI could help clinicians navigate language barriers and spend more time with patients.

Healthcare is feeling the strain. Multilingual AI can be the answer. 

Ten questions to Stuart Wood, Director of Product, on the economics of speech at scale, how close local accuracy now gets to cloud, and who could benefit from Speechmatics On-Device.

On-device speech-to-text, in 10 questions

Learn how law firms can reduce deposition costs with AI transcription by saving time, lowering expenses, improving efficiency, and streamlining workflows.

How Law Firms Can Reduce Deposition Costs with AI Transcription

Latest

Quantization was the key to fitting a cloud-grade model on a laptop. Getting the full optimization chain to cooperate around it was the hard part.

Speechmatics versus Whisper: how Adobe Premiere's on-device speech engine got rebuilt

Adobe Premiere users can run the most accurate on-device transcription locally; efficient enough for a laptop, powerful enough for professional work.

Adobe and Speechmatics deliver cloud-grade speech recognition on-device for Premiere

The metrics, milestones, and real-world outcomes that defined Voice AI’s breakout year.

Speechmatics in 2025: the numbers that shaped Voice AI's breakthrough year (+ what’s to come in 2026) 

Agent STT, powered by Linden 1, gives voice agents the speed, accuracy and conversational context they need in production.

Introducing Agent STT: built for the words that cost you the call

Scale more easily with Speechmatics 

Linden, Speechmatics' new speech-to-text model built for voice agents, is now available through LiveKit Inference, no separate API key, account, or invoice required.

Speechmatics & LiveKit Inference target the accuracy gap that breaks voice agents in production

Speechmatics is now live on Zapier. Connect industry-leading speech-to-text to 8,000+ apps with no code.

Speechmatics on Zapier: No-Code Speech-to-Text Automation

With 93% accuracy, our new model is twice as good as the nearest competitor.

Speechmatics launches Medical Model for real-time clinical transcription

Because in the real world, conversations are messy, and Voice AI needs to keep up.

Speaker Focus: Fixing Voice AI for the real world

A multilingual speech-to-text model from Speechmatics, with code-switching across [<langcount>] languages. Available today in production preview, starting with batch transcription.

Introducing Melia, our new multilingual speech-to-text model

A guide for voice AI engineers, ecommerce platforms and warehouse teams on SKU recognition accuracy voice assistant deployments depend on: why speech recognition systems produce transcription errors on product codes, what to measure when error rates matter, and the fixes that move the needle on order picking, voice ordering and customer-facing voice AI.

Alphanumeric speech recognition: why voice assistants mangle SKUs (and how to fix it)

A new integration gives agent developers a QA layer built for the complexity of the real world.

Speechmatics and Cekura bring real-world STT testing to voice agent pipelines

On Arabic and English, Melia 1 runs at less than half the mixed error rate of the next best model. On Mandarin and Tamil it switches more accurately than anything else we tested.

Melia 1 leads speech-to-text code-switching in Arabic, Mandarin and Tamil

An hour of audio in, a transcript back in under 20 seconds. 30 minutes in under 10 seconds. 10 minutes in under 5 seconds.

Gone in 20 seconds: Melia outpaces AssemblyAI, ElevenLabs and Soniox

Behind the build: what it takes to make cloud-grade speech recognition work inside Adobe Premiere, and why Whisper raised the stakes.

The Adobe story: How we made cloud-grade AI work on your laptop

Voice agents that pass in demos routinely fail in production. This guide covers the 11 best voice agent testing platforms in 2026, with the Five-Layer Testing Framework, platform deep dives, open-source alternatives, and a decision guide by maturity stage.

De-risk your voice agent: The 11 best voice agent testing platforms in 2026

End customers will only ever see their own provider's brand, which means the layer underneath has to be good enough to go unnoticed.

Tresic selects Speechmatics to power the speech layer of its conversation intelligence platform

How Wellcom Health uses real-time transcription, medical-grade accuracy and diarization to turn messy Dutch consultations into validated clinical reports.

Dutch doctors spend a quarter of their day on admin. Wellcom has built the fix.

Speech-to-text has moved from novelty to enterprise infrastructure. Here's how the leading platforms stack up in 2026 — and how to pick the right one.

Best speech-to-text AI guide: APIs, platforms and services compared

The court reporter shortage is reshaping litigation. Explore data, causes, and how legal teams are using digital reporting and AI transcription to adapt.

The court reporter shortage crisis: data, causes, and what legal teams are doing about it

Word error rate for legal transcription has no single acceptable threshold. But knowing how accuracy, audio quality, and review obligations connect to real legal risk is what separates a reliable transcript from a costly one.

What Word Error Rate Is Acceptable for Legal Transcription?

Faceless YouTube channels are becoming a scalable, system-driven way for businesses to produce high-impact video without relying on on-camera talent.