The University of Salford and TP Transcription Limited KTP 24_25 R3

Transcription services currently charge high prices for human-reviewed accuracy, while cheap automated options produce error-ridden text. This project builds new AI software that transcribes audio and video recordings with far better accuracy than existing low-cost tools, without the expense of human proofreaders. The problem is straightforward: businesses, journalists, researchers, and public sector organisations routinely need written transcripts of meetings, interviews, or lectures. High-quality transcription costs roughly £1–£2 per minute of audio, putting it out of reach for many. Cheap automated services often mangle technical terms, accents, or overlapping speech, making the output unreliable. This project targets that gap—the lower end of the market where price matters but so does usable accuracy. If the AI succeeds, it could make accurate transcription affordable for small businesses, local government, charities, and freelance professionals who currently go without. The technology would also open new income streams for the company involved, by offering a product that competes on both cost and quality. The work is applied commercial development rather than fundamental science, with a clear route to market through the existing transcription industry.