Podcast Transcription

Why it helps

Transcripts make episodes searchable on your site and accessible to deaf and hard-of-hearing listeners. At 6¢ an hour of audio, there's no reason to skip them.

Step by step

1. Export the final mixed episode as MP3, WAV or M4A.
2. Drop it into the audio workspace with text-only output.
3. Use the timestamped transcript to write chapters, and paste the plain transcript on the episode page.

Tips

• Process the final mix rather than separate tracks, so sponsor reads and intros are included in order.
• Crosstalk transcribes less accurately — edit those moments rather than the whole episode.

Example

Cost: 64 credits (6.4¢).

Limitations to know

• Hosts and guests aren't labelled — add names by hand where it matters.
• Music beds may be transcribed as fragments if they contain lyrics.

What each episode's transcript turns into

Timestamped transcript

Chapter markers for podcast apps and video uploads, and a quick way to find a quote by time.

Plain transcript

The full text on the episode page, for people who prefer reading and for search engines.

SRT or VTT

Captions for video clips and full-episode video uploads.

Pulled passages

Quotes for social posts and newsletters; see transcript to social media posts.

Show notes sit between these: a few lines on what the episode covers, the guest's links and the chapter list. Show notes from a transcript walks through writing them from the timestamped file.

Putting the speakers back in

mydubly doesn't mark who is speaking, so the transcript is one continuous text with no names attached. For a two-person interview, a fast way to restore them:

1. Search for the host's recurring phrases, like the welcome, "my guest today" and the sign-off, and mark those as the host.
2. Questions usually belong to the host and the answers that follow to the guest; scan down marking the start of each turn.
3. Use the timestamps to listen to the few passages where the turn-taking isn't obvious.
4. For panel episodes, consider naming people only in quoted passages rather than throughout.

The reasons software finds this hard are explained in why tools struggle to tell voices apart.

Working through a back catalogue

Work newest-first: recent episodes often get the most traffic and usually have your best audio. Name files by episode number so the transcripts line up with your feed, and split any live-show recording longer than two hours before upload.

Early episodes recorded on a laptop microphone or in an echoey room will need more correction than recent ones. Rather than polishing every old transcript to the same standard, decide what each one is for: a transcript that only makes the archive searchable can stay rough, while one you publish on a popular episode page deserves a full read.

A quick review pass before anything goes public

You rarely need to proofread an episode line by line. Errors cluster in predictable places, so check those first:

• Every guest name, company and book title mentioned on air; search the text for each one.
• Figures, dates and prices, which change meaning if a single digit is wrong.
• The stretches where people laughed or talked over each other; jump to them with the timestamps and listen.
• The first and last minutes, where music, cold opens and sign-offs sit next to speech.

Proofreading an AI transcript has a fuller checklist and tips for listening at speed.

Publishing the transcript on the episode page

A raw transcript is a wall of text. Before posting it, break it into paragraphs at topic changes, add the chapter titles as subheadings, put names at the start of each turn, and correct guest names and links mentioned on air. Place the transcript below the player and show notes, or behind a "Read the transcript" link, so the page still leads with the episode. If your hosting platform lets you attach a transcript file to each episode, the corrected plain text can go there as well, so listeners in podcast apps can read along. Publishing transcripts on your website covers layout and accessibility details.

Audio problems that show up as text problems

• Remote guests on a poor connection produce dropped or invented words exactly where the audio broke up; remote interview recording covers how to avoid it.
• Laughing and talking over each other can merge two people's words into one sentence, which is worth untangling only in passages you plan to quote.
• Intro and outro music with lyrics may leave fragments of the song in the text; delete them.
• Ads inserted dynamically by your host aren't in the master file, so they won't be in a transcript made from it.

For MP3 specifics see MP3 to text; to release the show in another language, see podcast translation.

Using the tool on this page

• The spoken language is detected automatically; you select the output language — the spoken language itself for an untranslated transcript, or another language for a translation. Files can be up to 2 hours long.
• Audio is sent over HTTPS in small chunks and deleted with its transcripts within 30 minutes of delivery (how files are protected).
• You can download a timestamped transcript, subtitles (SRT and VTT) and plain text. Every output, step and limit is explained on the audio to text page.