Player not loading? Watch on YouTube
This tutorial uses Buzz, a desktop wrapper for OpenAI's open source Whisper model, to transcribe a video into subtitles. The speaker demonstrates installation on Windows and lists macOS and Linux as supported platforms. Buzz provides a graphical interface for selecting a file and configuring transcription.
The walkthrough uses the medium model and English as the language. Model sizes shown include tiny, base, small, medium, large and large version three turbo. The speaker suggests starting with tiny if you lack a GPU; these are recommendations rather than measured hardware requirements. Buzz downloads a missing model automatically, so the offline transcription workflow requires the model to be available first.
The speaker enables speech extraction, leaves word-level timings off and selects SRT output. TXT and VTT are also available, along with translation into English. The demonstrated transcription finishes in 1 minute and 34 seconds and saves beside the input file. That timing describes this example, not a general performance guarantee.
The resulting subtitles contain a brand-name error. The speaker uploads the SRT to ChatGPT for correction and also describes using Gemma in LM Studio with web access. This correction step needs internet access in the demonstrated workflow, even when a local AI model handles the text. Uploading subtitles to a cloud service is separate from the initial transcription on the computer.
