Prepare the recording and your references
Use a class recording you are permitted to process. Keep any slides or reading list separately: this tool transcribes spoken audio, not writing on a board or equations visible in a video. An exported audio file or recorded lecture video can be selected above. If you need format guidance, start with audio files or video audio tracks. A distant microphone and student questions from across the room can make parts difficult to recognize.
Allow time for a long class
Choose the spoken language and start with a representative recording before committing to a very long lecture. The browser prepares bounded audio windows, but the overall task still uses memory and processor time. Keep the tab open and avoid concurrent heavy jobs. The first model download is separate from transcription time. Arabic uses its dedicated model, with about 1.33 GB of weights; neither the general model selector nor the GPU choice changes that Arabic path.
Check terms, formulas and quoted sources
Compare technical words, dates and references with the course material while listening. If a lecturer says “x squared,” the transcript is not an equation parser and may not reproduce mathematical notation. A phrase such as “as shown here” needs its slide context; do not expect visual information to appear automatically. Add your own study annotations in the working document only after checking the recording. The accuracy review guide explains which errors to prioritize.
Build your study material from reviewed text
Download TXT for the readable working version, or copy it into your own study document. Summaries, quizzes and study plans are not generated. Keep JSON if another tool needs the original segment times; review SRT or VTT separately if you want a caption track. Timed exports do not receive corrections made to the readable document. Save before leaving the page, and respect course permissions when sharing the transcript with other students.