Jump to content

Draft:VoisIT

From Wikipedia, the free encyclopedia

VoisIT is the first ever Faroese-language automatic speech recognition (ASR) mobile application, and it was developed by Andras Eliassen at iSolveIT ApS, a Danish software company [1]. The app converts spoken Faroese into written text using AI models trained on public datasets developed under the Ravnur and Ravnursson speech technology projects. It is available for free on both iOS and Android platforms.

Launched in August 2023, VoisIT is one of the first consumer-oriented ASR tools for the Faroese language and has been downloaded over 15,000 times—representing more than a quarter of the Faroe Islands' population.

History

[edit]

VoisIT was conceived following the development of the Ravnur [2] project, and later the Ravnursson [3] project that used the BLARK resources from Ravnur. The app’s creator was personally inspired after attending an early demo of the Ravnur speech model that showed the first ever Faroese ASR for Faroese. This demo resonated well, due to a family member’s experience with sudden hearing loss[4]. The underlying technology became publicly accessible via Hugging Face in 2023, enabling the first prototype.

VoisIT was launched publicly in August 2023 [5] and rapidly gained traction within the Faroese-speaking community. It was developed and published by iSolveIT ApS, a software startup founded in February 2023.

Features

[edit]

VoisIT offers the following functionality:

  • Push-to-record ASR interface: Users tap a button to begin recording, and tap again to submit the audio for transcription.
  • Automatic punctuation and capitalization of output text.
  • Manual editing of transcripts within the app.
  • Text-to-speech (TTS) output in Faroese.
  • Faroese-to-foreign-language translation support (in beta).
  • Simple export and sharing of transcriptions via standard mobile apps.

The app does not support live or real-time transcription but instead processes audio clips after they are recorded.

Development and Technology

[edit]

VoisIT builds on ASR models trained on over 100 hours of Faroese-language audio from the Ravnur and Ravnursson corpora. This was an Open Source project, running from 2019 to 2022, which aimed to gather high quality recordings from Faroese speakers of different ages, dialects, etc. These datasets were collected, labeled, and published by the Ravnur team. Later, the smaller Ravnursson dataset was created as a cleaned and shortened version of Ravnur - this work was made by a team of researchers at the University of Reykjavik, which included Carlos Hernández Mena and Annika Simonsen. This team also trained the speech model which initially was used by VoisIT.

The models are hosted on the open-source AI platform Hugging Face, enabling public use. VoisIT uses a pipeline of AI models to perform:

  • Audio transcription (ASR),
  • Punctuation and formatting,
  • Optional TTS playback.

All processing from the VoisIT app is initiated via user interaction, making the app suitable for asynchronous usage.

Since its launch, VoisIT has been connected to 3 different ASR models, and the effort is still ongoing, to improve accuracy.

Adoption and Impact

[edit]

As of mid-2025, VoisIT has over 15,000 downloads across iOS and Android platforms. In a country with roughly 55,000 residents, this constitutes a notable rate of adoption.

The app is used by a wide variety of users:

  • Deaf and hard-of-hearing individuals for communication assistance.
  • Students with dyslexia or learning difficulties, especially in essay writing and school exams.
  • General users needing text transcriptions of spoken Faroese for everyday productivity.
  • Non-native speakers who use VoisIT to learn to speak and pronounce Faroese

VoisIT represents one of the most widely used examples of Faroese-language AI technology in practice.

Roadmap and Future Development

[edit]

Planned updates to the app include:

  • Improved Text-to-Speech synthesis quality.
  • Enhanced translation accuracy and language support.
  • A chat interface allowing users to interact with one another using speech technology.
  • Broader multilingual support beyond Faroese.
  • Deeper accessibility integrations across platforms.

The development team continues to explore ways to strengthen support for under-resourced languages and speech communities.

[edit]

References

[edit]
  1. ^ "VoisIT talukennari". Máltøknidepilin. Retrieved 2025-06-24.
  2. ^ Admin (2022-07-04). "Verkætlanin Ravnur komin undir land". Máltøknidepilin (in Faroese). Retrieved 2025-06-24.
  3. ^ Hernández Mena, Carlos Daniel; Simonsen, Annika (2022-09-26). "Ravnursson Faroese Speech and Transcripts". {{cite journal}}: Cite journal requires |journal= (help)
  4. ^ "Verfaðir gav íblástur til talukennara". Kringvarp Føroya (in Faroese). Retrieved 2025-06-24.
  5. ^ "Nýggjur talukennari kann hjálpa deyvum og blindum". Kringvarp Føroya (in Faroese). Retrieved 2025-06-24.