Local AI in Practice: Transcribe YouTube with KB-Whisper
This is incredibly interesting for us. The cloud is good of course, but running AI locally on your own computer gives you full control over data, privacy, and costs. Even though it may seem a bit technically complex right now, development is progressing explosively fast. Starting with this today gives you and your school an enormous head start. We will go through how you can, step by step, download audio from YouTube and transcribe it to Swedish text with high quality. We use KB-Whisper, which is the Royal Library's Swedish language model. You will learn the basics of the terminal, manage Python environments, and run advanced AI models. These are skills that are becoming increasingly important in a digitalized school, so it is clearly worth getting started.
Moment i denna kurs
Understanding why running AI on your own computer is smart and getting to know the Swedish language model from KB.
Installing and configuring yt-dlp to automatically download audio from YouTube as MP3 files.
Create an isolated Python environment and install the libraries required to run advanced AI models for speech recognition.
Run the KB-Whisper model to convert your audio files to raw text and then use a second AI model to add punctuation and capitalization.
Use the powerful tool Pandoc to convert your clean text file into professional document formats like Markdown, DOCX, and HTML.