Podcast workflow
Transcribe a podcast episode locally
Create an editable podcast transcript, show-note source or timed caption file without uploading an unreleased episode to a transcription service.
Quick answer
Whisper Online downloads a whisper.cpp model, processes the selected media locally, lets you correct individual timed words, and exports the result without uploading the media file.
Step by step
The practical workflow
- 1
Choose the episode audio
Select the final mix or a speech-focused export from the device.
- 2
Run a suitable model
Start with Tiny or Base and use an English-only variant for English speech.
- 3
Review names and terminology
Correct hosts, guests, brands and specialist vocabulary in the editable transcript.
- 4
Export for publishing
Use TXT for an article draft, SRT or VTT for video versions, or JSON for custom tooling.
From recording to searchable episode text
A podcast transcript can improve accessibility, make an episode easier to scan and provide source material for descriptions, quotes and articles.
Word-level timing helps retain a connection between the written text and the recording. The detailed editor can correct both the wording and the time range of individual tokens.
Prepare audio for a better first pass
Clear dialogue, consistent levels and limited background noise generally produce a more useful draft. A final dialogue mix is often easier to transcribe than a multitrack production session.
- Review proper names and uncommon terminology manually.
- Expect overlapping speakers to require correction.
- Use shorter source sections on devices with limited memory.
- Keep speaker labels separately because diarization is not currently included.
Choose the export for the publishing task
TXT is suitable for a readable transcript or show-note draft. SRT and VTT are suitable for video podcast captions. JSON preserves each edited token's timing for developers and custom publishing systems.
Why use local transcription?
FAQ
Common questions
Can it label podcast speakers?
No. The current transcription produces timed words but does not perform speaker diarization.
Can I transcribe an unpublished episode privately?
The audio remains on the device. The browser downloads the model but does not upload the selected episode.
Can I use the transcript for video captions?
Yes. Export SRT or VTT after reviewing the timed words.
Which podcast formats work?
Common browser-readable audio formats such as WAV, MP3 and M4A are appropriate. Actual decoding support varies by browser.
Related guides