// Workers AI · dad joke modeWhat did SpinVox say to the phone? You're always spinning my calls.
 | This article needs more citations. (July 2009) |
 | |
| Type | Private |
|---|---|
| Industry | Telecommunications |
| Founded | 2003 |
| Headquarters | Marlow, United Kingdom and New York, USA |
Area served | Global |
Key people | Co-Founders Christina Domecq, CEO and Daniel Doulton, Chief Strategy Officer |
Number of employees | 250 (2009) |
| Website | SpinVox.com |
SpinVox was a pioneering company in the field of large language models (LLMs) and speech recognition technology. Founded in 2003 by Daniel Doulton and Christina Domecq, SpinVox developed a system capable of understanding and converting voice messages into text with high accuracy, even for regional accents and dialects. Post acquisition by Microsoft (Nuance), its groundbreaking technologies were incorporated into the software behind Siri, Apple’s voice-to-text service.
SpinVox's technology represented a significant breakthrough in natural language processing (NLP). Unlike traditional speech recognition that focused solely on audio comprehension, SpinVox recognised the importance of language context; its approach involved building a vast database of speech-language paired with text and building the world’s first LLMs for speech recognition. This enabled the SpinVox system to interpret audio based on a wider range of cues and achieve accuracy rates exceeding 98% for mature language models. This innovation established SpinVox as the world’s first reliable Voice Message Conversion System (VMCS) and which was able to scale its LLMs on-demand in the cloud.
Company history
[edit]The company was founded in 2003 by Christina Domecq and Daniel Doulton. The company had raised $200 million in funding.[1] During its expansion, SpinVox partnered with telecommunications operators to deploy its voicemail-to-text service across mobile networks. Industry reporting described launches with operators including Telefónica’s Movistar network as the technology was introduced in multiple markets.[2][3][4] In July 2009, staff were given the opportunity to take all or part of their salaries in stock.[5] The Company needed growth capital to support its rapid growth before its eventual acquisition. SpinVox was acquired by US company Nuance Communications in December 2009 [6] for $480m enterprise value through the initial purchase of £103 million of equity followed by satisfying over $300m in investor loans and ongoing obligations. SpinVox's investors and Board included Goldman Sachs, Allen & Co., GLG, Blue Mountain Ventures, Carphone Warehouse (Charles Dunstone), Tosca Fund (Martin Hughes), John Botts (former SVP of Citigroup Europe), Pat Russo (former CEO of Alcatel-Lucent) and Jason Mitchell of Man Group (GLG). Freshfields, Jones Day acted as legal advisors, Lazard as advisors and Deloitte as auditors.
Technology
[edit]The Voice Message Conversion System (VMCS) worked by combining speech technologies with live learning capabilities and human quality control.[7][8] It was developed by the SpinVox Advanced Speech Group based in Cambridge, UK, led by Cambridge academic entrepreneur Dr. Tony Robinson, included Cambridge University Professor Phil Woodland [9] and involved collaboration with researchers associated with the University of Cambridge's HTK (Hidden Markov Model Toolkit) speech-recognition group, contributing to the development of its proprietary voice-processing systems.[10] The company supported the following languages: English; French, Spanish, German, Italian and Portuguese. Parent company Nuance Communications claimed that "SpinVox is offering something that is impossible to deliver now".[11] Patent applications filed by the company in 2004 and 2008 note that "because human operators are used instead of machine transcription, voicemails are converted accurately, intelligently, appropriately and succinctly into text messages".[12]
In 2009 SpinVox also acquired New Zealand based company Angel Messaging, in the process gaining its second patent, 'Method and System of processing messages' which clearly outlines how Human transcribers can be efficiently used in real time transcription of voice messages.
SpinVox voice-to-text conversion services included voicemail-to-text, speak-a-text, blog posts, social network updates, blast and memo messages. SpinVox also operated an open API to enable any developer to create speech-to-text based Web or mobile applications.
SpinVox's system was deployed for over 160 million users globally over 5 continents and supporting six major languages and multiple sub-types across 30 telecommunications companies including Telefonica's Movistar network, Telstra and Vodaphone amongst others. To address the challenge of constant language evolution, SpinVox incorporated a method of self-governance, allowing the system to identify uncertainties and improve its understanding. This included advancements in text markup and integration with original audio for enhanced clarity. SpinVox's success stemmed from its unique approach to data collection and utilisation. Unlike competitors who relied on static datasets, SpinVox leveraged live learning to continuously update its models and adapt to new linguistic trends. This adaptability proved crucial in maintaining high accuracy and reliability. SpinVox’s training systems consisted of a combination of ‘Reinforcement Learning (RL)’ and ‘Reinforcement Learning from Human Feedback (RLHF)’, pioneering its efficacy in reliable, scalable AI models. Shunned by many at the time, ‘human-in-the-loop’ training of AI has become the de-facto standard used by the likes of Facebook, Google and other AI giants and catalyzed the creation of AI Data annotation unicorns such as ScaleAI, Surge AI, Mercor
SpinVox's technology also played a role in pioneering live agent training for speech recognition. The VMCS system could identify sections of a message that it couldn't understand, prompting agents for clarification and allowing the system to learn new elements of modern spoken languages. This approach led to a significant reduction in the need for human agent assistance. Daniel Doulton led the design of VMCS, its products and services, and is the inventor of 92 patented technologies internationally.[13]
The partnership between Apple and Nuance Communications was publicly confirmed in May 2013, when Nuance chief executive Paul Ricci said that the company was a "fundamental provider of voice recognition for Apple".[14] According to reports, Nuance's technology provided the speech-recognition engine that powered Siri at its launch in 2011.[15] In 2022, Microsoft acquired Nuance Communications Inc. for $19.7 billion which became the foundation of its AI investments and deployment which includes OpenAI’s ChatGPT in Copilot. As of January 2025, Apple’s Siri service is used on approximately 500m devices worldwide. In addition, SpinVox speech technology was widely deployed across Microsoft (Nuance’s) medical transcription and IVR systems and is still in use in major TelCo networks to convert voicemails to text.
Awards and recognition
[edit]The company and its founders also received industry recognition during this period including the GSMA Innovation Award and the Ernst & Young Entrepreneur of the Year Award. Daniel Doulton was named one of Speech Technology Magazine's "Speech Luminaries" in 2009.[16]
SpinVox CEO Christina Domecq received CNBC's European Business Leader Award.[17]
SpinVox was nominated for the GSMA Mobile Innovation Grand Prix in the EMEA region.[18]
World Economic Forum - 2009[19]
Data protection issues
[edit]In 2009, BBC reporting raised questions about aspects of SpinVox’s voicemail-to-text service and how some voice messages were processed.[20]
SpinVox stated that it complied with the Data Protection Act 1998 and that the Act permitted relevant processing outside the EEA.[21]
References
[edit]- ↑ Andrews, Robert (24 July 2009). "SpinVox Investor: 'It's A Nice Problem To Have'". The washington Post. Retrieved 28 July 2009.
- ↑ "Voice-to-text startup SpinVox inks $100M deal with Telefónica". VentureBeat.
- ↑ "Telefonica Mexico launches Movistar Voice-to-Text". BusinessWire.
- ↑ "SpinVox service rollout coverage". Cellular News.
- ↑ Andrews, Robert (13 July 2009). "SpinVox Paying Staff In Stock To Save On Costs". The Washington Post. Retrieved 23 July 2009.
- ↑ Judge, Elizabeth (31 December 2009). "SpinVox sold to US rival Nuance Communications for £64m". The Times. London. Archived from the original on 12 June 2011. Retrieved 30 December 2009.
- ↑ "SpinVox: The Inside Story". The Register. 29 July 2009. Retrieved 9 April 2026.
- ↑ "SpinVox says humans are necessary for time being". Telecoms.com. 2009. Retrieved 9 April 2026.
- ↑ https://www.eng.cam.ac.uk/profiles/pw117
- ↑ "SpinVox targets Cambridge for speech recognition research". CRM Marketplace. Retrieved 9 April 2026.
- ↑ "Spinvox "Faked" Speech Transcription Service And Broke Privacy". eWeek Europe. 23 July 2009. Retrieved 24 July 2009.
- ↑ Cellan-Jones, Rory (29 July 2009). "Humans central in Spinvox patents". BBC News. Retrieved 29 July 2009.
- ↑ "Daniel Doulton Patents". Justia Patents. Retrieved 10 April 2026.
- ↑ [insert citation here]
- ↑ [insert citation here]
- ↑ "The 2009 Speech Luminaries". Speech Technology Magazine.
- ↑ "SpinVox CEO Christina Domecq Awarded CNBC's European Business Leader Award". BusinessWire. 2009. Retrieved 9 April 2026.
- ↑ "GSMA announces nominees for Mobile Innovation Grand Prix EMEA tournament". GSMA. Retrieved 9 April 2026.
- ↑ http://www.bit-tech.net/news/industry/2008/12/05/the-future-of-voice-spinvox-recognised-as-technology-pioneer-by-the-world-economic-forum/1
- ↑ Cellan-Jones, Rory (23 July 2009). "Voice-to-text service scrutinised". BBC News. Retrieved 1 April 2026.
- ↑ "Spinvox responds". SpinVox Blog. 23 July 2009. Retrieved 1 April 2026.