Podcast Feed Transcript & Mention Extractor

Turn direct podcast feeds and publisher episode pages into structured episode records. The Actor extracts RSS, Atom, RDF, iTunes, and Podcasting 2.0 metadata; downloads publisher-provided transcripts; and finds timestamped literal keyword mentions.

No API account or external catalog credentials are required. The Actor does not transcribe audio and does not claim to search everything spoken across all podcasts.

What data can the Actor extract?

Field	Description
inputUrl	Original feed or publisher-page URL
resolvedInputType	direct_feed or publisher_episode
feed	Published feed GUID, title, URL, site, author, categories, and language
episode	Published GUID, title, page, audio, release date, duration, description, season, episode number, and image
peopleStatus	Whether people came from episode tags, channel fallback, or were missing
people	Explicit Podcasting 2.0 name, role, group, avatar, and profile/contact URL
transcriptStatus	Available, missing, fetch failed, too large, unsupported, or parse failed
transcriptSources	Every transcript descriptor published for the episode
transcript	Selected source, detected format, language, and optional text or cues
keywordMentions	Keyword, source timing, cue ID, precision, context snippet, and audio fragment
coverageReason	Plain-language explanation for unavailable coverage
sourceUrls	Feed, transcript, episode, and audio source URLs
scrapedAt	ISO timestamp when the record was emitted

Supported inputs

• Direct RSS, Atom, and RDF podcast feeds
• Publisher episode pages whose HTML advertises an RSS or Atom feed and whose canonical episode URL matches exactly one feed item

Catalog and contextless media URLs are intentionally unsupported:

• Podcast Index, Apple Podcasts, and Spotify catalog URLs
• Search terms, people searches, and category discovery
• Direct audio files
• Standalone transcript files without episode context

Use the podcast publisher's RSS/Atom feed URL instead of a catalog page.

Input parameters

Parameter	Description	Default
startUrls	Up to 200 direct feed or publisher episode URLs	[]
maxItems	Global dataset item cap	100
keywords	Up to 100 literal transcript terms	[]
maxEpisodesPerFeed	Newest records accepted from one direct feed	20
publishedAfter	Optional ISO 8601 publication-date lower bound	unset
includeFullTranscript	Include normalized transcript text	false
includeSegments	Include every normalized transcript cue	false
maxTranscriptBytes	Per-transcript decompressed byte cap	5242880
maxMentionsPerEpisode	Maximum keyword matches per episode	100
maxTotalTranscriptBytes	Global decompressed transcript byte budget	104857600
maxRequests	Global HTTP request budget	2000
runTimeoutSeconds	Run deadline in seconds	900
maxConcurrency	Global concurrent request cap	10
proxyConfiguration	Optional publisher-host proxy configuration	disabled
debugMode	Store bounded sanitized failure artifacts	false

An empty startUrls array finishes successfully with an empty dataset. Malformed, unsupported, or unresolved individual URLs are skipped with warnings so other submitted feeds can continue.

Examples

Extract recent feed episodes

{
  "startUrls": [{ "url": "https://feeds.podcastindex.org/pc20.xml" }],
  "maxItems": 5,
  "maxEpisodesPerFeed": 5
}

Find transcript mentions

{
  "startUrls": [{ "url": "https://feeds.podcastindex.org/pc20.xml" }],
  "keywords": ["Spotify", "value for value"],
  "includeSegments": true,
  "maxItems": 3
}

Process multiple feeds within strict budgets

{
  "startUrls": [
    { "url": "https://feeds.podcastindex.org/pc20.xml" },
    { "url": "https://example.com/podcast.xml" }
  ],
  "maxItems": 10,
  "maxEpisodesPerFeed": 5,
  "maxRequests": 100,
  "maxTotalTranscriptBytes": 52428800
}

Output example

{
  "inputUrl": "https://example.com/feed.xml",
  "resolvedInputType": "direct_feed",
  "feed": {
    "guid": "example-podcast-guid",
    "title": "Example Podcast",
    "url": "https://example.com/feed.xml",
    "siteUrl": "https://example.com/podcast",
    "author": "Example Network",
    "categories": ["Technology"],
    "language": "en"
  },
  "episode": {
    "guid": "episode-42",
    "title": "A Practical Example",
    "pageUrl": "https://example.com/podcast/episode-42",
    "audioUrl": "https://cdn.example.com/episode-42.mp3",
    "releaseDate": "2026-08-20T12:00:00.000Z",
    "durationSeconds": 2700
  },
  "peopleStatus": "episode",
  "people": [{ "name": "Alex Example", "role": "host", "group": "cast" }],
  "transcriptStatus": "available",
  "keywordMentions": [{
    "keyword": "example",
    "startTimeSeconds": 123.45,
    "timestamp": "00:02:03.450",
    "precision": "cue",
    "snippet": "Here is a practical example for the discussion."
  }],
  "coverageReason": null,
  "sourceUrls": {
    "feed": "https://example.com/feed.xml",
    "transcript": "https://example.com/episode-42.vtt",
    "episode": "https://example.com/podcast/episode-42",
    "audio": "https://cdn.example.com/episode-42.mp3"
  },
  "scrapedAt": "2026-09-01T12:00:00.000Z"
}

Transcript and mention coverage

Transcript coverage is publisher-dependent. The Actor supports VTT, SRT, JSON, HTML, and plain-text transcripts declared in Podcasting 2.0 feed metadata. Untimed transcript text can produce a literal match but cannot produce a truthful timestamp. Full transcript text and cue arrays are disabled by default.

People are returned only from explicit Podcasting 2.0 metadata; the Actor does not guess guests from titles or descriptions.

Pricing

The Actor uses pay-per-event billing and charges exactly one event per emitted episode. Episodes with a successfully normalized transcript use the transcript-enriched-episode tier; all other emitted episodes use the episode-metadata tier. Unresolved inputs, retries, and duplicates are not charged. Current event prices are shown in the Apify Console before a run starts.

Responsible use

Publishers retain rights in their feeds, transcripts, and audio. Enable full transcript redistribution only when you have a lawful basis and appropriate permission. You are responsible for complying with applicable privacy, copyright, and database laws.

Support

Use the Issues tab for bugs or feature requests and the API tab for programmatic runs.