Skip to main content

Transcribing Many Episodes

You can transcribe a whole back catalogue in one go, and see what it will cost before anything is queued. Bulk transcription uses your own provider account, so every episode is charged to you at your provider’s rate.

If you already have transcripts from another service, importing them is free.

Starting a Run

On Benecaster → Episodes, on a show with Transcription switched on, you have two ways to start:

  • Transcribe selected is in the bulk menu. Tick the episodes you want, choose it and click Apply.
  • Transcribe all without transcripts is a button in the toolbar, and needs no ticks. It covers every page of the list, drafts included, because a migrated back catalogue often arrives as drafts. It leaves out the trash.

An episode you’ve already transcribed from its Transcript tab, on any of its audio, is skipped.

The Confirmation

Either way, a panel opens in the toolbar and nothing is queued until you click Transcribe N episodes. It reads Transcribe N episodes with {provider}?, followed by:

  • The estimate: Estimated cost: $X, charged to your {provider} account., with the minutes, the rate per minute and the date of the list price it comes from.
  • What is skipped: episodes that already have a transcript, episodes already queued or processing, and episodes with no audio file. Finished transcripts can’t be transcribed again in bulk.

The figure is an estimate. Benecaster can’t see your invoice, a negotiated rate or a volume discount, so your provider’s bill may differ. Where a provider’s price depends on something the request doesn’t fix, the estimate uses the higher rate, so it doesn’t read low.

A few things can change what the panel says:

  • An episode with no duration is still queued and charged. The panel says its cost is not in the figure.
  • A custom provider with no list price gets no estimate. The panel says so and tells you to check your provider’s pricing.
  • With OpenAI connected, an episode whose recorded file size is over 25 MB is not queued, counted or priced. The panel says Skipping N episodes over OpenAI Whisper’s 25 MB upload limit. They would only fail, so they are not queued or priced. To transcribe them, use another provider. An episode with no recorded size is still queued, and fails at no charge if it turns out to be too large. See Choosing a Provider.
  • Audio on another site. The panel says N of these have their audio on another site. and If your host uses dynamic ad insertion, the transcribed audio may include ads. A CDN in front of your media library counts as another site.
  • A paused queue. The panel says the episodes will wait until you click Resume queue.

Which Audio?

When an episode in the run has a tier’s own file, the confirmation asks Which audio?:

  • Base audio, the default.
  • Best available: each episode’s highest tier with its own file, or its base audio when it has none.

It then shows {n} using base audio, {n} using a tier’s own audio. and works the cost out again. An episode transcribed from a tier’s file shows no transcript on the episode page, and its lower tiers get none in their feed, until its base audio has one. The confirmation warns you of that, which is why Base audio is the default. A tier’s file with no recorded length is priced at the base audio’s length, and the confirmation says so.

A failed episode is tried again and charged again the next time you include it.

Cancel on Transcribe selected prints Nothing was queued. and clears your ticks.

Watching the Queue

Each episode’s line in the list shows its state and updates by itself. Episodes go to your provider no faster than Max concurrent transcription jobs allows.

While two or more episodes are waiting, the notifications bell shows Transcription queue: 47 of 100 remaining. It links to Settings → Transcription. Dismissing it lasts for the run.

Pausing and Resuming

While something is queued or the queue is paused, Settings → Transcription opens with a Transcription queue section, with Pause queue or Resume queue.

Pausing covers all your shows and sends nothing new. A job your provider already has still finishes and is charged. Format transcript is not affected. While paused, the bell reads Transcription queue paused: … whatever the count, even if you had dismissed the notice.

What You’ve Spent

Settings → Transcription shows an estimate of the month so far: see Transcribed This Month. Benecaster doesn’t set a spending limit, and nothing stops at a figure, because every run starts with a confirmation showing its cost.

For Developers

See Also

Need this built rather than just documented? See our services →

Last updated