About Speechmatics
Speechmatics is a technology company that develops automatic speech recognition software. It provides multilingual speech-to-text, text-to-speech and voice AI technology for enterprises, developers, and platform partners. Its primary products are speech-to-text, covering real-time and batch transcription, and text-to-speech. The technology can be deployed on-premises and in public or private cloud.
Speechmatics is headquartered in Cambridge, United Kingdom, and is a private company. It was originally named Cantab Research Ltd when it was founded in 2006 by speech recognition specialist Dr. Tony Robinson. Katy Wigdahl is its Chief Executive Officer.
In 2019, Speechmatics raised £6.35 million in a Series A funding round with investment from Albion Venture Capital, IQ Capital, and Amadeus Capital Partners. In July 2024, it released Flow, an API for voice interactions.
Drafted from en.wikipedia.org · not yet confirmed by Speechmatics
Open jobs
From Speechmatics's own job board.
Nothing open at Speechmatics right now.
Their tool
In Tools & Services. Listed, not recommended.
Recently
From Speechmatics's own newsroom feed and from the hiring team.
Distributed Self-DistillationFrom their feed
In this blog we delve into how we can effectively use FSDP in self distillation training, where there is both a student teacher present.
October 14, 2025
How to build smarter turn detection for Voice AIFrom their feed
A guide to how to use a Small Language Model to improve Turn Detection in voice agents
May 12, 2025
How to Finetune Sesame AI's Speech Model on New Languages and VoicesFrom their feed
A comprehensive guide to finetuning Sesame AI's CSM on new datasets.
April 25, 2025
Exploring the trade-off between speed and accuracy in real-time transcriptionFrom their feed
Complete this sentence:
September 6, 2024
Sparse All-Reduce in PyTorchFrom their feed
The All-Reduce collective is ubiquitous in distributed training, but is currently not supported for sparse CUDA tensors in PyTorch. In the first part of this blog we contrast the existing alternatives available in the Gloo/NCCL backends. In the second part we implement our own efficient sparse All-Reduce collective using PyTorch and CUDA.
March 16, 2024