How to transcribe a podcast
Automatic speech recognition is a fast first draft. It is less dependable when speakers overlap, names are unusual or a recording contains music. The publishing step is a human review, not a download button.
Explore the guideThe short answer
Use a clean recording, generate a transcript in the right language, correct it while listening, label speakers and export as readable text or timed captions depending on the destination.
From recording to publishable text
Preserve source audio
Keep your original recording. If you have separate speaker tracks, retain them for easier review.
Create the draft
Select the language and process a short sample first. Note where background music or crosstalk causes errors.
Listen and correct
Focus first on names, figures, negations and speaker identities. Then adjust punctuation for reading without changing meaning.
Export for use
Create a readable web transcript, subtitle file or quote sheet. Verify timecodes in the destination player.
One missing word matters
“We do not recommend doubling the dose” becomes unsafe if the system drops “not.” Prioritize critical claims over cosmetic punctuation changes.
Choose the right output
Full transcript
Useful for accessibility, research and searching the conversation.
Captions
Need timed segments, line breaks and on-screen review.
Show notes
Summarize the episode for listeners; they are not a substitute for a transcript.
Frequently asked questions
+How accurate is AI transcription?
Accuracy varies with recording quality, accents and subject matter. Test your own audio and count meaningful errors.
+What is diarization?
Assigning segments of speech to different speakers; automatic labels still need correction.
+Can Spotify display transcripts?
Spotify documents transcript management for creators, with availability and options that can vary by account.
Your next idea can be an episode
Turn the guide into a first draft. Review the result before sharing it.
Create my podcast