Setting Up Transcription

Transcription Service sends your episodes to your own account with a transcription provider. This page covers connecting that account and the settings that shape every transcript. They are at Benecaster → Settings → Transcription, and one provider and API key serve all your shows.

The Transcription settings appear while Transcription Service is switched on for at least one show (Benecaster → Settings → Add-ons → the Transcription Service card → Active on).

Choosing a Provider

OpenAI receives the audio file itself and refuses anything over 25 MB. The other three fetch the episode from its address. Benecaster downloads the audio first and, when it is too large for OpenAI, fails the job with This episode’s audio is larger than OpenAI Whisper’s 25 MB upload limit. before anything is sent or charged. For episodes longer than about 25 minutes, choose AssemblyAI, Deepgram or Rev.ai.

Every provider labels speakers only for a show with Label speakers on, in that show’s Transcription settings tab. It’s off until you turn it on.

Costs go straight to your provider at their published rates. Each provider’s own site has its current prices.

Transcribed This Month

Above Provider, the page shows Transcribed this month: N minutes, about $X., with the note An estimate from each provider’s list price as of {date}. Your provider’s bill may differ. It counts finished transcriptions only, not imported transcripts, failed jobs or Format transcript.

It covers all your shows and follows your site’s timezone. Minutes from a provider with no list price are counted, and the line says That figure leaves out N minutes from a provider with no list price. With only such minutes there is no dollar figure. OpenAI is priced at the model set now, because the model isn’t stored with each transcript. This is an estimate to keep an eye on, not a limit: nothing stops at a figure.

While episodes are queued, or the queue is paused, a Transcription queue section sits at the very top of the page, with Pause queue or Resume queue. See Transcribing Many Episodes.

Connecting Your Provider

1. Go to Benecaster → Settings → Transcription.
2. Under Provider, choose OpenAI Whisper, AssemblyAI, Deepgram or Rev.ai.
3. For OpenAI, choose the OpenAI model: Whisper (original), or gpt-4o-transcribe-diarize (labels speakers). The note under it reads Only gpt-4o-transcribe-diarize labels speakers, and only for a show with Label speakers turned on in that show’s Transcription settings tab. It is off until you turn it on.
4. Paste your provider’s key into API key.
5. Click Test connection. On success it reads Connected. {Provider} accepted this key. Save settings to keep it. Testing is free with every provider and transcribes nothing.
6. Click Save settings.

How Your Key Is Kept

• A key your provider rejects is never saved, and the error names the provider.
• A key that could not be checked, because the provider was unreachable or busy, is saved but shown as Saved, but not checked yet. Use Test connection. until a test succeeds.
• Once saved, the key is never shown again, not even its last characters. The field shows a row of dots and Replace, and the status line reads Connected. Last checked {date}. To change the key, click Replace and enter a new one.
• Changing provider asks for that provider’s key: Enter your {Provider} API key. Your current key belongs to another provider and will be replaced.
• If Benecaster can no longer read the key, which can happen after your site’s encryption key changes, the status line reads Benecaster can no longer read your saved API key, so it needs to be entered again. Enter it again and save.

The key is stored encrypted on your site and is never sent to benecaster.com.

The Setup Reminder

Until a provider has accepted a saved key, the dashboard’s Needs attention list shows Connect your transcription provider. with Add API key →, which opens these settings. It goes away by itself once Test connection or Save settings confirms the key, and comes back if a later test finds the key rejected, or Benecaster can no longer read it. It shows on every show’s dashboard while Transcription is switched on for at least one show. You can dismiss it for your own account with ×.

Language

Language starts on Auto-detect, which lets your provider work out each episode’s language. You can instead choose from the languages your provider supports, listed by name in your admin language. Choosing the language your episodes are in can improve accuracy. It’s also what is recorded as the episode’s transcript Language when a transcription finishes; on Auto-detect, the language your provider reports is recorded instead.

Rev.ai cannot detect the language. On Auto-detect it transcribes as English and records what Rev.ai reports, normally English, so choose the language your episodes are in. If you switch to a provider that doesn’t offer the language you chose, the setting goes back to Auto-detect.

Max Concurrent Transcription Jobs

How many episodes are with your provider at once, from 1 to 5. It starts at 1. Higher values finish a batch of episodes faster but make your provider’s rate limit likelier. Under the setting, the page shows what your chosen provider usually handles, and the line changes as soon as you pick another provider:

It is advice only; any value from 1 to 5 can be saved. A change applies to the next episode sent. When your provider says too many requests are arriving, Benecaster waits and retries, and never lowers this setting for you; see How Episodes Wait Their Turn.

Vocabulary Hints and Context

A text box, below Max concurrent transcription jobs, sent with every transcription so your provider spells your show’s words right. Up to 1,000 characters.

List the names, places and words your show uses often, separated by commas: Benecaster, Ana Ruiz, Saskatoon, kombucha. How each provider reads it:

• OpenAI Whisper reads the whole text as context for each episode.
• AssemblyAI, Deepgram and Rev.ai read only short terms separated by commas or new lines. A sentence reaches only OpenAI Whisper.

The box starts with Welcome to the podcast. This is a clearly punctuated transcript of a conversation between the host and their guests. That sentence helps OpenAI Whisper punctuate, and the other three providers ignore it. Reset to default, shown while the text differs from it, puts it back. Save with Save settings.

HTML Formatting

Below Vocabulary hints and context. Format transcript, in an episode’s Transcript tab, sends a finished transcript to OpenAI, charged to your OpenAI account and typically under $0.01 per episode.

• Formatting instructions are what OpenAI is told, up to 4,000 characters. Reset to default puts the original back.
• OpenAI API key (for transcript formatting) appears only when your provider is AssemblyAI, Deepgram, Rev.ai or a custom one. With OpenAI as your provider, its key is reused. The key is checked with OpenAI for free when you click Save settings, a refused key is not saved, and it is stored encrypted like your provider key. Once saved it shows as a mask with Replace and Remove. Remove asks first: Remove your OpenAI API key for transcript formatting? Format transcript becomes unavailable while your provider is not OpenAI. Transcripts you have already formatted stay as they are., with Remove key and Keep key. A key Benecaster can no longer read offers Remove too.

The whole page saves with one Save settings.

The Setup Wizard Step

When Benecaster’s setup wizard runs while Transcription is switched on for at least one show, a step headed Set up transcription. appears just before the last step, with Provider, API key and Test connection, and links to each provider’s pricing page.

Continue stays unavailable until Test connection passes. The test is free and stores nothing, and a key that passes is then saved, shared by every show. A refused key shows the provider’s reason and saves nothing. A site whose key was already accepted finds Continue available at once, with Connected. Your {provider} API key is already saved.; editing either field makes it test again. To skip, choose I’ll set this up later in Settings.

If you finished setup before switching Transcription on, Benecaster invites you instead: An add-on has a new setup step for {show}., listing Set up transcription, with Set it up and No thanks. It asks once, on your lowest-numbered show with Transcription on, and only while no key has been accepted. Every answer is final. A key your provider accepts on Settings → Transcription removes the invitation straight away and for good. A key saved but unchecked, because the provider could not be reached, leaves it up until Test connection passes, and a refused key leaves it up. See Setup Steps From Add-ons You Install Later.

These settings are always available here, on Settings → Transcription, whatever you answered.

Next Steps

• Choose who can read each show’s transcripts: Transcript Visibility. A show starts on All visitors.
• Transcribe your first episode from its Transcript tab or from Benecaster → Episodes: Transcription Overview.
• Or import a transcript you already have. That needs no key.