Exporting and Formatting Transcripts
A finished transcript can leave your site as a file, for you and for your listeners. You can also have OpenAI lay it out in paragraphs, which changes how it looks on the page and in the PDF.
Exporting a Transcript
Open the episode’s Transcript tab. Under Imported. N lines. or Transcribed by {provider}. N lines. is Export:, with one button per format. Each downloads {episode-slug}-transcript.{ext}, or {episode-slug}-{tier}-transcript.{ext} for a tier’s own transcript, made from the transcript as it is now, your corrections included. It costs nothing and calls no provider. The buttons are hidden while Edit transcript is open.
| Format | Keeps times | Names speakers | Good for |
|---|---|---|---|
| SRT | Yes | Yes: every cue opens with [Speaker 1]: |
Video captions |
| VTT | Yes | Yes: <v Speaker 1>…</v> |
Web players |
| Plain text | No | Yes: a new block, after a blank line, opens with Speaker 1: where the speaker changes |
Show notes |
| No | Yes: Speaker 1: opens the line where the speaker changes, or in bold once formatted |
Sharing and printing |
Speakers are named when your provider detected them, which takes speaker labels turned on for the show. A transcript with no speakers exports without names.
A plain-text import has no times to export, so it offers Plain text and PDF only.
About the PDF
The PDF is A4: the episode title, then the show name and publication date above a rule, then the transcript’s words, with Page n of m at the foot. A timed transcript prints one paragraph per line, with the speaker named where it changes, and a plain-text import keeps its own paragraphs. Once a transcript is formatted, the PDF uses the formatted version, with paragraphs and speaker names in bold. The first download of a long transcript takes a moment, and later ones come from a saved copy.
PDF is offered only for Latin, Greek and Cyrillic letters. A transcript, episode title or show name in Chinese, Japanese, Korean, Arabic, Hebrew, Hindi, Thai or another script gets no PDF. The tab says PDF is offered only for transcripts written in Latin, Greek or Cyrillic letters, which is what it can print. SRT, VTT and plain text work for every language., and visitors simply see no PDF link. A transcript with speakers also gets no PDF when your site’s language translates Speaker N into a script the PDF can’t print, even if the transcript itself is in Latin letters.
It needs the PHP extensions dom and mbstring. Most hosts have both. Without them the tab says PDF is not available on this server: it needs the PHP extensions dom and mbstring.
Saved copies hold the transcript’s text. They are kept in wp-content/uploads/benecaster-transcription/pdf/ under random names, are refused where the server reads .htaccess, and are removed when the plugin is deleted. Reset Plugin Data can’t remove them; they stay until the plugin is deleted or the transcript changes.
Downloads for Your Listeners
Under the transcript on the episode page, and in [benecaster_transcript] in full and scrollable modes, visitors who may read it get Download the transcript: with the same formats. Visitors your transcript visibility excludes get no links, whatever the teaser shows them. Under Hidden, only people who can edit the episode see them.
These links are the same public files a podcast app is given, so anyone who has the address can open them. See Transcript Files Are Public. The links stay when your license lapses or Transcription is switched off for the show.
A developer can remove some or all of the links: see Choose which transcript downloads your listeners get.
Formatting a Transcript
Format transcript sends a finished transcript to OpenAI, which lays it out in paragraphs with speaker names in bold. Your words are kept. It runs only when you click it.
The button is in the episode’s Transcript tab under Export:, with its cost beside it: This sends your transcript text to the OpenAI API and incurs a small additional charge (typically under $0.01 per episode). The charge goes to your own OpenAI account. It runs in the background (Formatting with OpenAI. A long episode can take a few minutes; you can leave this page.), and the tab updates every 15 seconds.
When it finishes, the button reads Format again and the tab says Formatted. The episode page, the PDF and podcast apps that read HTML transcripts now use the formatted version. Correcting the transcript removes it, so format again after corrections. If it fails, OpenAI’s reason shows in red and nothing is kept.
It needs an OpenAI key. With OpenAI as your transcription provider, its key is reused. With AssemblyAI, Deepgram, Rev.ai or a custom provider, add a separate one on Settings → Transcription. Until there is a key the button is greyed out, with the reason and Open Settings → Transcription.
What Formatting Changes
- The episode page and
[benecaster_transcript]show formatted paragraphs, speaker names in bold, with timestamps still clickable and the spoken line still highlighted. Each paragraph opens with its first line’s timestamp. If OpenAI’s layout can’t be matched back to every line, the page keeps the timestamped lines.mode="preview"always shows the first timed lines. - The PDF uses the formatted version.
- Podcast apps that read HTML transcripts get it through a second
<podcast:transcript>tag. See Podcasting 2.0 Support.
It follows the same visibility as everything else. SRT, VTT, plain-text exports and site search are not affected.
Correcting the transcript, or replacing it, removes the formatted version. Format it again afterwards, which is another small charge.
For Developers
See Also
Need this built rather than just documented? See our services →
Last updated