Trint takes audio/video of interviews, speeches, news conferences, meetings and turns around transcripts of 90%+ accuracy in less than half an hour. Automated speech-to-text (S2T) software continues to improve, but it makes mistakes and no sensible person would rely on it. Trint is the first software to create a method of fast and failsafe verification opening up vast new markets for automated S2T software. In minutes users can search, verify, edit and export the content. |
|
Trint is a cloud-based software that allows users to upload audio files (MP3, WAV, etc.) or low resolution video files (MP4, MOV etc.). In minutes Trint returns automated transcripts, but these are not the familiar error-laden text documents. A Trint tethers the original words (and video) to the text on the screen – click on a word, hear the word. Its patented innovation uses the algorithm’s confidence values to show users the words that are most likely incorrect. Users can tab to the low confidence words, hear them and either accept them or correct them. Long interviews, speeches, meetings are instantly searchable and verifiable. Trint’s simple drag-and-drop edit system makes export easy.
Amidst all the technological innovation in media, transcribing the content of interviews, speeches, etc., remains stuck in the 1970’s. It is still done manually. The growth of contextually-aware natural language programming (NLP) enabling computers to derive meaning from human or natural language input is revolutionising speech-to-text technology (S2T). Until recently, however, error rates have made S2T of little value to journalists and others. Increasing levels of accuracy mean ST2 now has the potential to be a useful tool, but only if it can be trusted and verified. Trint is the first software to provide a reliable system for verifying and correcting the inaccuracies of S2T software.
Trint has low hourly costs for transcription and computing. B2B customers will be charged per seat, with a data cap. The strategy is to make the core service of Trint so affordable that it is not considered a luxury. Focused innovation will allow charging more: recognition of key speakers; a dictionary of technical terms for each vertical; security and storage. Once the UX is refined, Trint will be offered in other languages and will be opened to the B2C market, where freelancers and consumers can buy packages of transcription hours.
With an untapped market overwhelmed by video, podcasts, recorded talk there is a huge appetite for Trint’s savings of time and labour. Trint aims to become the language of manipulating talk. We will establish our lead in B2B, then take our software global in other languages and to B2C users. The potential margins are 5x and 10x cost and the volumes in the four verticals can reach millions of hours. We anticipate revenue in Y3 of £5M+, positioning ourselves for sale in the £20-£40M range.