On this page

WIPO Speech to Text

WIPO Speech to Text is a powerful transcription tool which automatically converts audio and video content into text using artificial intelligence. It was specifically designed for international meetings and conferences. Once trained in a specific subject area, it out-performs other general purpose transcription tools.

The software was originally created and deployed to help transcribe official WIPO meetings and can be customized for other organisations worldwide.

Contact us to find out more
(Photo: Jaouad.K/Getty Images)

In practice: Find out how the Speech to Text  tool is used at WIPO

What are the benefits of WIPO Speech to Text?

Performance

The tool can transcribe one hour of video/audio in five minutes when supported by the appropriate IT infrastructure.

Accuracy

The tool uses the latest neural machine learning technology, and is particularly effective in transcribing audio from non-native speakers.  

Security

WIPO Speech to Text can be installed on-site to guarantee security and confidentiality.

What do I need to use it?

  • A GPU-based server infrastructure if you wish to run the tool independently. If not, cloud-based infrastructure will work.
  • For best results, a large collection of domain-specific data in the relevant language can be used to customize the tool.

Which languages does it work with?

WIPO Speech to Text is currently only available in English. We are working on extending the tool to work with other five official United Nations languages (Arabic, Chinese, French, Russian and Spanish).

How can I get it for my organization?

WIPO Speech to Text is available via standard licensing agreements and WIPO's team of experts can even help you to install and set up the tool.

Send us a mail to find out more.

Resources

Our user guide gives you a quick walkthrough of how to use our WIPO Speech to Text tool.

Who uses it?

WIPO Speech to Text tool is used by organizations hosting international meetings to help them to accurately and efficiently transcribe the proceedings.

Testimonials

  • UNOG’s project FAST has been looking for an automatic speech recognition solution that would be accurate for the UN context with its multilingual voices, diverse accents and specialized lingo, and has found the perfect match in WIPO’s Speech to Text tool.

    In October 2019, it started running the customized WIPO Speech to Text for public calendar meetings, servicing its iconic clients, such as OHCHR, ODA and UNCTAD, and received overwhelmingly positive feedback from secretariats and delegates alike.

    We are looking forward to further collaborating with WIPO, including for other official working languages.

    Sofia Lobanova Zick, FAST Project Manager, United Nations Office at Geneva

  • CERN's Digital Memory project has digitized its massive back catalogue of audiovisual material. Faced with a lack of accurate text descriptions for much of the content, the project turned to audio-to-text transcription technologies that could provide full-text indexes to the entire collection. Combined with initiatives from CERN's Translation Service and Diversity Office - also keen to provide greater accessibility to historical content – CERN chose to collaborate with WIPO during the summer of 2019 to carry out an audiovisual transcription pilot project. The WIPO Speech-to-Text software, which was designed for conferences involving foreign English speakers with a wide variety of accents, was used to process a sample of historic recordings of High-Energy Physics seminars.

    Despite the originality and complexity of the vocabulary, very encouraging results have been achieved, providing a strong incentive to set up a generalized transcription service open to the entire CERN audio-visual community, to process both the old data and the on-going events and content.

    Jean-Yves Le Meur, Head of Digital Memory Project, CERN

Artificial intelligence

WIPO Magazine

Knowledge

Cooperation and policy