If you work with educational video and audio, you probably spend more time than you would like creating transcripts and subtitles. This is where I have found Whisper particularly useful. Whisper is an AI speech recognition system developed by OpenAI. Give it an audio or video recording and it converts the spoken words into text. It supports many languages and copes surprisingly well with different accents, background noise and specialist terminology. It can also translate speech from various languages into English. For me, the most obvious application is subtitling. I can take a recorded lecture, interview, webinar or educational video and use Whisper to produce the first version of the subtitles. Instead of typing everything manually, I only need to check and correct the result. This saves a considerable amount of time. But transcription is only the beginning. Once I have the text, I can use AI tools to create a summary, chapter titles, keywords, a short description for a learning platform or even questions for learners. For a one-hour lecture, for example, Whisper can provide the raw transcript that becomes the starting point for several useful learning resources. Another interesting application is multilingual production. We increasingly work with recordings in languages that we do not speak ourselves, like West-Frisian, Irish and Basque in a recent production of expert interviews. Whisper can help us understand and edit these recordings and prepare them for translation and subtitling. Whisper is open source, which also means that it can be installed locally. This can be particularly interesting in education when recordings should not be uploaded to an external service.
It is not perfect, and I would never publish its output without checking it (or having it checked by native speakers. Names, technical terms and poor recordings can still cause problems. But as a first step in turning spoken educational content into usable text, I found Whisper remarkably practical.
