Transcribe episodes

Turn an episode into a transcript note. For a podcast you added, when its feed lists a transcript in a format that PodNotes reads, PodNotes saves that one, with no API key and at no cost. Otherwise, PodNotes transcribes the audio with your own OpenAI or Deepgram account.

Transcribe an episode

PodNotes: Transcribe current episode first uses the transcript the episode’s feed publishes, at no cost, for a podcast you added to PodNotes. For any other episode, and when the feed publishes none in a format that PodNotes reads, it transcribes the audio with your provider and API key, and your provider bills you for it. Without your provider’s key on the device, the command never bills you. It saves a transcript that the feed publishes. For an episode without one, it shows a notice that asks for the key.

1. Play an audio episode in PodNotes. You can pause it.
2. Run PodNotes: Transcribe current episode from the command palette (Ctrl+P, or Command+P on macOS).

The transcribe commands need a connection, so they are missing from the command palette while the device is offline.

To transcribe audio, set up an OpenAI key. For speaker labels from Deepgram, set up a Deepgram key instead.

A notice shows the progress, and you can keep listening. When the transcript is ready, PodNotes saves it as a note and opens it, as Where a note opens describes. With the default settings, the transcript of the Acquired episode “The Home Depot” is transcripts/Acquired/The Home Depot.md.

Transcribe the audio instead

To transcribe the audio even when the feed publishes a transcript:

1. If this device has no key for your provider, set one up. See Set up an OpenAI key, or Add speaker labels for Deepgram.
2. Rename, move, or delete any transcript note you saved for the episode. PodNotes does not replace an existing transcript.
3. Run PodNotes: Transcribe current episode from audio.

The command always uses your provider and your API key. A transcript of the audio has paragraphs without times, so its lines do not link to their moment in the episode. With speaker labels, each turn starts with a label such as A: with OpenAI or 1: with Deepgram.

Read the transcript in the player

The full player’s Transcript panel shows the transcript that the episode’s feed publishes, for free, without saving a note. It reads Feed transcript and Free at the top.

1. In the full player, select Transcript.
2. Select a line with a time to play the episode from that moment. A transcript without times can only be read.

The panel highlights the line being spoken and follows it. To search the transcript, select Find in transcript. To keep the transcript, select Save as note. It saves the same note as PodNotes: Transcribe current episode, but never transcribes the audio. Where the panel is narrow, such as in the sidebar or on a phone, the button shows only its icon, a page with a plus sign. Once the note exists, the button reads Open note and opens it. The note opens as Where a note opens describes. See Read the transcript for every control.

This clip opens Transcript, selects a line to play from it, and finds the word “walk”:

When the feed publishes no transcript, the panel says “No transcript in this feed”. For an audio episode, it offers Transcribe with OpenAI, or Transcribe with Deepgram when your speaker labels come from Deepgram, with a Paid tag. It works like PodNotes: Transcribe current episode from audio, with your provider and your API key. When the episode’s transcript note already exists, the panel offers Open note in place of that button. Offline, the panel’s transcribe button and Save as note are turned off, and Open note still works.

Transcripts that feeds publish

Some feeds list a transcript for each episode with the Podcasting 2.0 <podcast:transcript> tag. PodNotes uses it for a podcast you added, while the feed still lists the episode.

• When an episode lists several transcripts, PodNotes picks WebVTT, then SRT, then JSON, then HTML, then plain text.
• In a transcript with times, each paragraph starts with the time it begins. The time uses the time format in Timestamp template. With the default format, - {{linktime}} with a space at the end, the time is a link that plays the episode from that moment. With {{time}}, it is plain text. PodNotes ignores the rest of the format, such as the - list marker.
• When the transcript names its speakers, each speaker’s turn starts with Speaker label format, even with Speaker diarization off.
• A plain text transcript, or an HTML transcript without times, has paragraphs but no times.

With the default settings, the Markdown of a transcript with times and speakers looks like this:

[00:00:00](obsidian://podnotes?...) **Ben:** Welcome back to the show. Today we look at how the company started.

[00:00:41](obsidian://podnotes?...) **David:** It started with two founders and one store.

Set up an OpenAI key

To transcribe the audio without speaker labels, PodNotes sends it to OpenAI with your API key. A transcript that the feed publishes needs no key.

1. Create a key on the OpenAI API keys page.
2. In Obsidian, open Settings → PodNotes.
3. Under Transcripts, select the secret picker next to OpenAI API key.
4. Create a secret that holds your key, or select one that you saved before.

On each other device, select or create the secret again. See API keys and private feeds on each device.

Add speaker labels

To label who speaks in a transcript of the audio:

1. Open Settings → PodNotes.
2. Under Transcripts, turn on Speaker diarization.
3. In Diarization provider, select OpenAI (gpt-4o-transcribe-diarize) or Deepgram.
4. If you selected Deepgram, create a key in the Deepgram console.
5. If you selected Deepgram, under Deepgram API key, select or create a secret that holds the Deepgram key. With OpenAI, PodNotes uses your OpenAI key.

The providers differ in these ways:

Provider	Key	Labels	Long episodes
OpenAI	Your OpenAI key	A, B, C	Sent in parts of about 20 MB. A in one part may be B in the next.
Deepgram	A Deepgram key. No OpenAI key needed.	1, 2, 3	Sent in one request. Labels stay the same for the whole episode.

A transcript that the feed publishes uses the speaker names in the feed, with or without Speaker diarization.

Change the speaker label format

1. In Settings → PodNotes, under Transcripts, turn on Speaker diarization. Speaker label format, below it, becomes available.
2. Set Speaker label format, for example to **Speaker {{speaker}}:** with a space at the end.

The format sets the text before each turn. The default, **{{speaker}}:** with a space at the end, starts a turn with A: in bold. With **Speaker {{speaker}}:**, a turn starts with Speaker A:. With Deepgram, the labels are numbers, as in Speaker 1:. The labeled text fills {{transcript}}, so Transcript template needs no change.

Transcripts that feeds publish use the format too. If you turn Speaker diarization off again afterwards, they keep using it.

Change where transcripts go and what they contain

Transcript path under Transcripts sets where the note goes, transcripts/{{podcast}}/{{title}}.md by default. Keep .md at the end, because PodNotes does not add it. Transcript template sets the content, and {{transcript}} is the transcript text. Transcript and timestamp tags lists every tag and the default template.

What transcription costs

A transcript that the feed publishes costs nothing. To transcribe the audio, PodNotes charges nothing, and your provider bills your account per minute of audio:

• OpenAI API pricing lists the rate for plain transcription (whisper-1) and for speaker labels (gpt-4o-transcribe-diarize).
• Deepgram pricing lists the rate for pre-recorded audio with speaker labels.

Limits and error messages

These cases stop a transcription or change its result:

• PodNotes transcribes audio episodes only. For a video episode, both commands are missing. The Transcript panel can still show and save a video’s transcript that the feed publishes.
• Both commands need a connection, and they are missing while the device is offline.
• PodNotes transcribes the audio instead of using the feed’s transcript in these cases:
  • The episode is a local file.
  • You have not added the podcast to PodNotes.
  • The feed no longer lists the episode, or lists no transcript in a format PodNotes reads.
  • PodNotes cannot fetch the feed, and no copy of the feed on this device lists the episode.
• When PodNotes would transcribe the audio but the provider’s key is not on this device, it transcribes nothing, and a notice asks you to select or create the key in the transcript settings.
• When PodNotes cannot fetch or read the feed’s transcript, a notice starts with “The published transcript could not be fetched” or “The published transcript could not be read.” PodNotes does not transcribe the audio on its own. If your provider’s key is on this device, the notice names Transcribe current episode from audio, which uses your API credit. To try the feed again, run PodNotes: Transcribe current episode again.
• PodNotes waits 30 seconds for a feed’s transcript and reads at most about 16 MB of it.
• When a transcript note already exists at the path, a notice gives its path and PodNotes does nothing else. To transcribe again, rename, move, or delete the note first.
• PodNotes sends long audio to OpenAI in parts of about 20 MB. In a transcript without speaker labels, a part that fails three times leaves the marker [Error transcribing chunk N]. With OpenAI speaker labels, the marker is [Error diarizing chunk N]. To retry, delete the note and run the command again. When every part fails, PodNotes saves no note.

API key storage

PodNotes keeps your API keys in Obsidian’s secret storage on each device, and only the name of each key syncs. On the first start after an upgrade from a version that kept plain-text keys in data.json, PodNotes moves them to secret storage and does not finish loading until the move succeeds. If it fails, restart PodNotes. The move cannot erase old copies in backups or settings exports, so replace a key at the provider if it may have leaked. See API keys and private feeds on each device.