š§ by @davehappyminion
Summarize Podcasts
Take a podcast episode, transcribe it, extract the most valuable insights, and deliver a structured summary the user can scan in 2 minutes instead of listening for 60.
1. Receive Input
Accept one of the following from the user:
- Podcast URL ā A direct link to an episode page (Apple Podcasts, Spotify, YouTube, podcast website, or direct MP3/audio URL)
- Audio file path ā A local file (mp3, m4a, wav, etc.)
- Episode name + show name ā If provided without a URL, search for the episode and confirm with the user before proceeding
Validate the input:
- Confirm the episode exists and is accessible
- Check audio duration ā warn the user if the episode is over 3 hours (transcription will take longer)
- Display the episode title, show name, and duration for confirmation
2. Transcribe Audio
Transcribe the full episode using the best available method:
- Check if a transcript is already published (many podcasts provide them on their website or via RSS)
- If not, use a speech-to-text service or local transcription tool
- Preserve speaker labels where possible (e.g., "Host:", "Guest:")
- Include timestamps at regular intervals (at minimum every 2-3 minutes)
If transcription fails, report the error clearly and suggest alternatives (e.g., trying a different URL format, or providing the audio file directly).
3. Identify Speakers
- Determine how many speakers are in the episode
- Label them by name if identifiable from the episode intro, show notes, or context
- If names cannot be determined, use "Speaker 1", "Speaker 2", etc.
- Note any guest introductions with their title/role if mentioned
4. Extract Key Segments
Analyze the transcript and identify:
- Key insights ā Novel ideas, surprising facts, actionable advice
- Notable quotes ā Memorable or shareable statements
- Disagreements or debates ā Points where speakers diverge
- Recommendations ā Books, tools, people, or resources mentioned
- Action items ā Anything the listener is encouraged to do
For each segment, record the timestamp so the user can jump directly to it.
5. Generate Summary
Produce a structured summary at three levels of detail:
- One-liner ā A single sentence capturing the episode's core message
- Key takeaways ā 5-8 bullet points covering the most important ideas
- Detailed notes ā Section-by-section breakdown with timestamps
6. Present Summary
Deliver the summary using this template:
# Podcast Summary
**Show:** [Show Name]
**Episode:** [Episode Title]
**Date:** [Publication Date]
**Duration:** [HH:MM:SS]
**Speakers:** [Speaker 1 (role)], [Speaker 2 (role)]
## TL;DR
[One sentence summary of the episode]
## Key Takeaways
1. [Takeaway with brief context] ā [TIMESTAMP]
2. [Takeaway] ā [TIMESTAMP]
3. [Takeaway] ā [TIMESTAMP]
[...up to 8]
## Notable Quotes
> "[Quote]" ā [Speaker], [TIMESTAMP]
> "[Quote]" ā [Speaker], [TIMESTAMP]
## Detailed Notes
### [Topic/Segment 1] ā [START TIMESTAMP]
[2-4 sentence summary of this segment]
### [Topic/Segment 2] ā [START TIMESTAMP]
[2-4 sentence summary]
[...continue for each major segment...]
## Mentioned Resources
- [Book/Tool/Person/Link] ā mentioned at [TIMESTAMP]
- [Book/Tool/Person/Link] ā mentioned at [TIMESTAMP]
## Worth a Full Listen?
[Brief honest assessment: is this episode worth listening to in full, or are the notes sufficient?
Note which segments are worth listening to directly if only parts are strong.]
7. Save and Offer Extras
- Save the summary to the user's configured output location
- Offer to save the full transcript as a separate file
- Offer to add any mentioned books/resources to the user's reading list
Safety Rules
- Never redistribute or publish the full transcript. It is for personal use only.
- Never fabricate quotes or timestamps. If a timestamp is uncertain, mark it as approximate with "~".
- Never editorialize the speakers' views. Summarize what was said, not what you think about it.
Notes
- This workflow is on-demand only. No schedule.
- Transcription quality varies by audio quality. If the source has heavy background music or crosstalk, note this in the summary.
- For video podcasts (YouTube), extract the audio track only. Do not attempt to describe visual content unless the user requests it.
- If the podcast has show notes with timestamps already, use those as a starting scaffold and enrich from there.
- Long episodes (2+ hours) may benefit from a "chapters" approach ā break the summary into logical sections with clear timestamp ranges.
