How to transcribe a podcast episode and turn it into show notes

The short answer

Three steps. Get the audio file. Run it through a transcription app. Reshape the text into whatever you actually need. The middle step is the easy one and takes a few minutes of waiting. The first and third steps are where people get stuck.

Sort one thing out before you start, because it decides the whole route: are you working from a file or from a link? A handful of browser tools will fetch the audio down from a YouTube or RSS address for you, and that only helps when the episode is posted on a platform the tool supports. Phone apps work the other way round. They transcribe a file that already exists on your device, in a format like M4A, MP3, WAV or AAC, and there is nowhere to paste a Spotify or Apple Podcasts address. Noter AI is one of those: files in, no link import. So if all you have is a link, downloading or exporting the episode is genuinely step one, and the next section covers how.

Here is what a finished transcript should give you, whichever tool you use:

• The full text, with a timestamp on every segment so you can find a moment later
• Speaker labels, so a two-host show or a host-plus-guest interview reads as a conversation and not a wall of text
• Synced playback, tap a line and hear that exact second of audio, which is how you check a quote before you publish it
• A summary and the key points, so you are not writing show notes from a blank page
• An editable transcript, because every episode has a name or a product term the AI will spell wrong
• Translation, if part of your audience reads in another language

Step 1: get the audio file

How you do this depends on whose episode it is.

If it is your own show, the file already exists. Export it from whatever you edit in. In Audacity use File > Export Audio, then choose MP3 as the format in the export window. In GarageBand on Mac use Share > Export Song to Disk and pick MP3 or AAC; GarageBand on iPhone and iPad shares a song out from the My Songs browser instead. In Descript, Riverside, Adobe Audition or Logic, look for Export and choose an audio-only format. Export the final mastered version, the one listeners actually hear, not the raw session. That matters more than it sounds, and the troubleshooting section explains why.

If it is someone else's show, check the episode page on the show's own website first. A lot of podcasts publish a direct MP3 download button, and many RSS feeds point straight at a plain MP3 file you can save. That is the clean route. Downloading an episode inside Apple Podcasts or Spotify for offline listening does not help, because those apps keep the file locked inside themselves and it never appears in the Files app.

One honest note on that. Transcribing an episode for your own reading or searching is normally fine; republishing someone else's full transcript as your own page is not. Quote it, link it, credit it.

1. 1Export or download the episode as M4A, MP3, WAV or AAC.
2. 2Get the file onto the phone you will transcribe from. AirDrop is fastest from a Mac to an iPhone. Otherwise drop it in iCloud Drive, Google Drive or Dropbox, then open that folder on the phone.
3. 3If you only have a video export of the episode, go back to your editor and export an audio-only file. Transcription apps for audio, Noter AI included, do not accept MP4 or MOV.
4. 4Check the file plays before you import it. A zero-byte or half-uploaded file is the most boring reason a transcription fails.
5. 5Stay on Wi-Fi for the import. A 60-minute episode is a real upload, and transcription happens in the cloud, not on the phone.

Step 2 on iPhone: import the file and transcribe it

Two routes, and both end in the same place. Use the Share sheet if the file is sitting in another app. Use the in-app importer if the file is in Files or iCloud Drive. The wider version of this workflow is in how to transcribe audio on iPhone.

1. 1Open the app holding the episode, for example the Files app or Voice Memos.
2. 2Long-press the file and tap Share, or tap the Share button in the top corner.
3. 3Scroll the app row in the share sheet and tap Noter AI. If you do not see it, tap More and turn it on once.
4. 4The file imports and transcription starts on its own. A progress indicator shows where it is.
5. 5For the in-app route instead, open Noter AI, tap + or New Note, then Import Audio, and pick the episode from the Files browser.
6. 6When it finishes, the transcript opens with timestamps on every segment and each voice labelled as a separate speaker.
7. 7Tap any line to hear that exact moment. Use this to check anything you plan to quote.

Step 2 on Android: the same job, different buttons

Android is close enough that the workflow reads the same, with different menu names. Files usually live in Files by Google or My Files, and sharing goes through the Android share sheet rather than the iOS one.

1. 1Install Noter AI from Google Play and sign in with the same account you use on any other device.
2. 2Open Files by Google (or your phone maker's file manager) and find the episode.
3. 3Tap the three-dot menu on the file, then Share, then Noter AI.
4. 4Or open the app first, tap New Note, then Import Audio, and browse to the file.
5. 5Wait for transcription to finish, then open the note.
6. 6Notes sync across devices, so an episode you import on an Android phone is readable on an iPad later without doing anything.

The other way: record the episode straight into your phone

If you record in a room, with a co-host across the table or a guest in the same space, you can skip the export step entirely and record into the transcription app itself. This is not how you make a broadcast master, so keep your proper mics running for the audio you publish. Treat the phone as a second capture that exists purely to produce text.

Recording keeps going with the screen locked, so a 90-minute session does not need you watching the phone. A Control Center widget starts a recording in one tap, which is useful when you remember 20 seconds after the guest started talking. If your episode is a remote interview on Zoom, Google Meet, Microsoft Teams or Webex, you can send a bot to join the call instead, and the same transcript comes back.

1. 1Put the phone flat on the table between the speakers, roughly equal distance from each.
2. 2Open Noter AI and tap New Note, then Start Recording. Or tap the Control Center widget and skip opening the app.
3. 3Say the names out loud in the first minute. It gives you an easy anchor for relabelling speakers later.
4. 4Lock the screen if you want. Recording continues.
5. 5Tap Stop when the session ends. Transcription starts by itself.
6. 6Rename the note to the episode title straight away, so it is findable when you have 40 of them.

Other ways to transcribe a podcast, and what each one costs you

An AI app is not the only route. Depending on what you need the text for, one of these may be enough.

• Just want to read one episode? The built-in transcript in Apple Podcasts or Spotify is free and already there. It is the right answer for a listener who is not going to do anything with the text.
• Want the text out of the app? That is where the free options run out. Apple lets a listener copy a short passage, capped at 200 words and stamped with a link back to the episode, and YouTube lets you copy out of the transcript panel. None of the in-app transcripts give you a file you can edit or publish in full, so anything you publish, search, or paste at length needs a real transcription tool.
• Publishing weekly? Per-minute pricing is what kills people here. Four hour-long episodes a month is 240 minutes of audio, every month, forever.

Turn the transcript into show notes and timestamped chapters

A raw transcript is not show notes. Nobody reads 9,000 words of speech. The useful move is to reshape the same recording into the formats you actually publish, and this is where the output styles earn their keep.

The point is that one transcript should be able to produce several different documents without you rewriting anything. Noter AI does that through six output styles, and the one that really matters to a podcaster is Custom instruction, where you describe the shape and voice you want in plain words. Show notes are a house style, so being able to state that style rather than pick from a menu is the whole trick.

Some prompts that produce something publishable on the first try:

1. 1Open the finished note and choose the Custom output style.
2. 2Ask for the show notes: "Write podcast show notes. Two-sentence hook, then five bullet takeaways, then a list of every book, tool, person and company mentioned."
3. 3Run it again for chapters: "List the topics in order as chapter headings, each with the timestamp where that topic starts." Every transcript segment carries a timestamp, so it has the raw material to work from.
4. 4Check the chapter timestamps against the transcript before you paste them into your host. Tap the line, hear the audio, confirm it lands in the right place.
5. 5Run it a third time for social: "Pull three quotable lines under 25 words each, with the speaker name and the timestamp."
6. 6Switch to the Email style if you send a newsletter with each episode, or Custom instruction if you want to spell out the voice so the notes sound like you rather than like an AI.
7. 7Export the finished text as PDF, Word (.docx) or plain text, or use rich-text copy to paste it straight into your podcast host's episode description field.

If your show is bilingual, or your listeners are not

This is the part most transcription tools handle badly. Plenty of shows are not recorded in one clean language. A host in Dubai flips between Arabic and English inside a single sentence. An Indian tech show runs in Hindi with English product names all the way through. A Miami business podcast moves between Spanish and English depending on who is talking.

Tools that make you pick a language before transcribing break on this. Whatever you choose, the other language comes back wrong: either transliterated phonetically into the language you selected, so English brand names turn into nonsense syllables, or dropped from the transcript entirely. You end up fixing it by hand, which defeats the point.

Noter AI transcribes in 60+ languages with automatic detection, and tags the language word by word rather than locking the whole file to one choice. A sentence that carries two languages arrives with each part in its own script. You never pick a language before importing. If your audience is spread across countries, you can also translate the finished summary into any of 18 languages: Arabic, Chinese (Simplified), Chinese (Traditional), Dutch, English, French, German, Indonesian, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Thai, Turkish and Vietnamese. Translating the summary is the useful move, not the whole transcript, because the summary is the part anyone reads. There is a longer walkthrough in how to translate meeting notes, and a full comparison of tools on this in the best app for transcribing multilingual meetings.

Troubleshooting: six things that actually go wrong

None of these are exotic. If your first episode came back wrong, it is probably one of the six.

• The file is too long, or the import stalls. Long episodes are big uploads and transcription runs in the cloud. Stay on Wi-Fi, keep the phone plugged in, and let it finish. If it still will not go through, re-export at a lower bitrate, 128 kbps mono is plenty for speech and can halve the file size. Failing that, split the episode into two or three parts in your editor, transcribe each, and keep them together in one folder.
• Both hosts came back as one speaker. This happens when everyone shared a single microphone, or when the two voices are close in pitch. The fix at the source is to record each host to their own track and transcribe the tracks separately. The fix after the fact is to edit the speaker labels and names in the transcript yourself, then re-run the summary so it picks up the corrected names. More on how the labelling works in can AI tell who said what in a meeting.
• Music beds produce junk text. An intro sting, a bed under the ad read, or a loud outro can come back as stray words, because the model is looking for speech and finds something that resembles it. Two fixes. Transcribe the pre-master version, before music was laid under it. Or trim the musical intro and outro off a copy of the file before importing, and paste your fixed intro copy in by hand afterwards.
• Chapter timestamps do not match the published episode. This one bites podcasters who transcribe the raw session. If you cut 90 seconds of throat-clearing at the top and three tangents in the middle, every timestamp after each cut is wrong for listeners. Always transcribe the final exported file, the exact audio you are going to publish. If you re-edit after transcribing, re-transcribe.
• Names, products and jargon are spelled wrong. Guest names, company names and technical terms are the usual casualties. Open the transcript, fix the spelling once, then re-run the summary so the corrected term flows through into your show notes rather than sitting fixed in one place and wrong in another. Detail in can you edit an AI transcript and re-run the summary.
• The remote guest sounds thin and comes back garbled. Compressed call audio is the worst input you can give a transcription model. Record the guest locally on their end and have them send you the file, the double-ender approach, then transcribe that track. If you only have the call recording, expect to correct more by hand and consider transcribing the guest's track separately from yours.

What it costs, and when doing it by hand stops making sense

If you publish once and never again, do it free. Copy the text out of YouTube's transcript panel, or run Whisper on your laptop, and spend an evening cleaning it up.

The maths change the moment the show is a habit. A weekly hour-long episode is around four hours of audio a month. Per-minute human transcription on that is a real line item. Per-seat AI tools charge for every person on the show, which is an odd way to price a two-host podcast where only one person does the admin. Otter.ai Pro is $16.99/month, Notta Pro is $13.99/month, and Fireflies runs $10 to $19 per seat per month. Rival prices checked September 2026 and they move often, so confirm on the vendor's own page before you commit.

Noter AI is $9.99/month or $49.99/year, flat. Not per seat, not per minute, not per episode. The same subscription covers your episodes, your interview prep, and the calls you take about the show. There is a free trial, so test it on a real episode before you decide.

You get the transcript with speaker labels and timestamps, synced playback for checking quotes, six output styles including a Custom instruction for show notes, an editable transcript that re-runs the summary after a fix, AI chat you can point at one episode or at your whole back catalogue at once, folders and search across every note's title and AI notes, and export to PDF, Word or plain text. Available on iPhone and iPad and on Android.

The back-catalogue search is the part podcasters underrate. Once 30 episodes are transcribed and sitting in one place, you can ask a question across all of them and find the episode where a guest said the thing you half-remember, in seconds, instead of scrubbing through audio.