How to Transcribe Any Podcast for AI (Apple Podcasts & RSS to Text)

2026-11-19来源 dev.toEnglish
How to Transcribe Any Podcast for AI (Apple Podcasts & RSS to Text)

TL;DR: Podcasts are one of the richest sources of expert knowledge, but LLMs can't listen to them. To use them in ChatGPT, Claude or a RAG pipeline you need transcripts. The reliable way: resolve the show's RSS feed, download each episode's audio, split long episodes into chunks, transcribe them with Whisper and return structured JSON with timestamps. Below you'll find a no-code option, JavaScript and Python examples, and how to transcribe new episodes automatically.

Thousands of hours of interviews, expert explanations and industry news are published as podcasts every week. Almost none of it is searchable, quotable or usable by AI until it's text.

This guide shows how to get clean, timestamped transcripts from Apple Podcasts links, RSS feeds and MP3 files, one episode or a whole back catalogue, without running any servers yourself.

Why transcribe podcasts?

  • RAG and AI assistants: let a chatbot answer questions from a show's entire archive, with timestamps to the exact moment something was said
  • Research and monitoring: track what experts, founders or competitors say across dozens of shows
  • Content repurposing: turn episodes into show notes, blog posts, newsletters and social posts
  • SEO and accessibility: publish full transcripts next to your own episodes

How podcast transcription works under the hood

  1. Find the RSS feed. Every public podcast has one. An Apple Podcasts link can be resolved to its feed through Apple's public lookup API.
  2. Pick the episodes: the newest N, everything published after a date, or one specific episode.
  3. Download the audio straight from the podcast host. No proxies or browsers needed: podcast audio is meant to be downloaded.
  4. Split long episodes. Speech-to-text APIs limit upload size, so a 2-hour episode is converted to compact mono audio and cut into 30-minute chunks, with timestamps shifted back afterwards.
  5. Transcribe with Whisper large-v3-turbo, which handles accents, crosstalk and 90+ languages well.
  6. Return structured JSON with the full text, timestamped segments and episode details.
  7. Optionally, let an LLM write the show notes: a summary, key takeaways, timestamped chapters, people mentioned and notable quotes.

You can build this with an RSS parser, ffmpeg and a Whisper API. If you'd rather skip the plumbing, the Podcast to Text Actor on Apify does exactly this, and I'll use it in the examples below:

Podcast Transcript API: Apple, RSS & MP3 to Text · Apify

Transcribe Apple Podcasts, RSS feeds and MP3s to text with timestamps, AI summaries and chapters. Whisper AI in 90+ languages, about $0.24 per hour.

favicon apify.com

Why Spotify links don't work

Spotify streams podcasts with DRM, so its audio can't be transcribed. The good news: almost every show on Spotify is also on Apple Podcasts or has a public RSS feed. Search for the show on Apple Podcasts and use that link instead.

Option 1: No code (2 minutes)

  1. Open Podcast to Text and click Try for free.
  2. Paste an Apple Podcasts link or an RSS feed. This one is real, so you can try it right away:
https://podcasts.apple.com/us/podcast/npr-news-now/id121493675 
  1. Set Episodes per podcast (for example 5 for the five newest) or Only episodes published after a date.
  2. Optional: turn on AI summary to also get a summary, chapters, key points and quotes, or pick a language in Translate transcript to.
  3. Click Start and download the transcripts as CSV, JSON or Excel, or send them to Google Sheets.

Option 2: JavaScript

import { ApifyClient } from 'apify-client'; const client = new ApifyClient({ token: process.env.APIFY_TOKEN }); const run = await client.actor('aemon/podcast-audio-to-text').call({ urls: ['https://podcasts.apple.com/us/podcast/npr-news-now/id121493675'], maxEpisodesPerPodcast: 3, subtitleFormat: 'srt', aiSummary: true, // summary, key points, chapters, people and quotes }); const { items } = await client.dataset(run.defaultDatasetId).listItems(); for (const episode of items) { if (episode.status === 'ok') console.log(episode.episodeTitle, episode.aiSummary?.chapters); else console.log('Failed:', episode.inputUrl, episode.errorCode); } 

Option 3: Python

from apify_client import ApifyClient client = ApifyClient("YOUR_APIFY_TOKEN") run = client.actor("aemon/podcast-audio-to-text").call( run_input={ "urls": ["https://feeds.npr.org/510318/podcast.xml"], "maxEpisodesPerPodcast": 10, "publishedAfter": "2026-11-01", } ) for episode in client.dataset(run["defaultDatasetId"]).iterate_items(): print(episode["episodeTitle"], episode["durationSec"], episode["text"][:200]) 

What the output looks like

Each episode becomes one item like this (values shortened):

{ "status": "ok", "podcastTitle": "Your Podcast", "podcastAuthor": "Jane Host", "episodeTitle": "Episode 42: How AI agents actually work", "episodeUrl": "https://example.com/episodes/42", "audioUrl": "https://cdn.example.com/episode-42.mp3", "publishedAt": "2026-11-12T08:00:00.000Z", "durationSec": 2731, "language": "en", "text": "Welcome back to the show. Today we're talking about...", "wordCount": 7480, "segments": [{ "start": 0, "end": 4.1, "text": "Welcome back to the show." }], "aiSummary": { "summary": "The hosts explain how AI agents plan, call tools and recover from mistakes, with examples from real products.", "keyPoints": ["Agents are loops of planning and tool calls", "Good tools matter more than bigger models"], "chapters": [{ "start": 0, "title": "Intro" }, { "start": 312, "title": "What an agent really is" }], "topics": ["ai agents", "llm", "tools"], "people": ["Jane Host"], "quotes": ["Good tools matter more than bigger models."] } } 

Episodes that are unavailable or skipped come back with "status": "error" and an error code, so you always know what happened to every link.

Transcribe new episodes automatically

To keep a knowledge base up to date:

  1. In Apify, create a Schedule that runs the Actor every morning with maxEpisodesPerPodcast: 1 for each show you follow.
  2. In the Integrations tab, send results to Google Sheets, Slack, a webhook, Make, Zapier or n8n.

Every new episode is transcribed and delivered without you touching anything. AI agents can also call the Actor through the Apify MCP server in Claude, ChatGPT or Cursor.

How much does it cost?

You pay $0.004 per minute of audio, about $0.24 per hour, and only for episodes that were transcribed successfully. A 30-minute episode costs $0.12. The optional AI summary with chapters costs $0.01 per episode and translation $0.003 per 1,000 words, and the Apify free plan's monthly credit covers roughly 20 hours of podcasts.

Transcribing short videos instead? The companion Video to Text Actor does the same for Instagram Reels, TikTok, X and YouTube.

Wrapping up

Podcasts hold a huge amount of knowledge that AI can't use until it's text. Resolve the feed, chunk long episodes, transcribe with Whisper and keep the output structured, and a whole back catalogue becomes searchable in minutes.

Which podcasts would you turn into a searchable knowledge base? Tell me in the comments. And if a feed doesn't work, open an issue on the Actor page.

Disclosure: I'm the author of the Podcast to Text Actor used in the examples, published on Apify under my agency Aemon.

原文:https://www.a8q9.cn/en/article/how-to-transcribe-any-podcast-for-ai-apple-podcasts-rss-to-text-b21d90.html

相关文章