How to Transcribe a Podcast With Deciphr (2026 Tutorial)
Deciphr turns a podcast episode into a transcript, a summary and show notes from a single link or file upload. A 45-minute episode typically processes in 5 to 15 minutes, depending on queue.
The pitch is repurposing, not transcription. You already have the audio; Deciphr gives you the text assets that make it findable and shareable — show notes, a blog-ready summary, social snippets.
What does Deciphr produce from one episode?
Five outputs: a full timestamped transcript, a chapter-style summary, show notes with guest details, a blog-post draft, and short social captions.
Output quality varies by type. The transcript is the most reliable. The blog draft is the weakest — it needs a rewrite before publishing, and usually a substantial one.
How do you process an episode?
- Create an account and click New Episode.
- Paste the RSS episode URL or upload an audio file (MP3, WAV, M4A).
- Add guest names and titles if you want the show notes to get them right — you can also let it infer, but inference on names is unreliable.
- Choose output language and, if available, the summary style.
- Submit and wait. RSS processing is usually faster than a straight upload because no transfer is needed.
- Review the transcript before using any downstream output.
Step 6 is non-negotiable. Every derived asset inherits transcript errors, so fixing names and jargon at the top of the chain fixes everything below it.
How accurate is the transcript?
On clean studio audio with two native speakers, error rates are low — typically under 5%, mostly punctuation and paragraph breaks rather than wrong words.
Accuracy drops on: heavy accents, crosstalk, phone-quality guest audio, and domain jargon. Technical podcasts suffer most. In a test transcript one product name appeared four different ways across the episode.
| Audio condition | Practical accuracy | Fix effort |
|---|---|---|
| Studio, 2 speakers | High | 5 min spot-fix |
| Remote guest on laptop mic | Fair | 15-20 min |
| 3+ speakers, overlap | Fair to low | 30+ min |
| Heavy accent + jargon | Low | Manual rewrite |
A useful trick: find-and-replace the three most common mis-spellings after the first episode, and save the list. Most shows have a stable vocabulary.
What do the free and paid tiers include?
There’s a free trial with a limited minute allowance — enough to process one episode and evaluate accuracy on your own audio. Paid plans are monthly with a minute or episode allowance that scales by tier.
Pricing is per-minute-ish rather than per-seat, which makes it cheap for a solo podcaster and reasonable for a small production team. Watch the allowance if you process every take, not just published episodes.
Where does Deciphr lose to alternatives?
If you only need a transcript, dedicated transcription tools are more accurate and cheaper per minute. Deciphr’s value is in the bundled summary and show notes, so paying extra only makes sense if you’d otherwise write those by hand.
If you need speaker-accurate diarization for legal or medical work, this isn’t the tool. It does speaker labels well enough for a podcast, not well enough for a compliance record.
And the blog draft output generally needs 30 to 45 minutes of editing per post. Treat it as an outline, not a draft.
Who is it for?
Podcasters publishing weekly, especially interview shows where show notes and guest links take real time. Also useful for anyone repurposing existing audio libraries into blog content.
Skip it if you have fewer than five episodes and don’t publish show notes — manual transcription plus a notes template is fine.
Quick Summary
- Deciphr produces a transcript, summary, show notes, blog draft and social captions from one episode link.
- A 45-minute episode processes in 5-15 minutes depending on queue.
- Accuracy is high on studio audio, poor on crosstalk and heavy jargon.
- Fix names and jargon in the transcript first — every derived asset inherits those errors.
- Treat the blog draft as an outline; it needs 30-45 minutes of editing.
