Podcast Transcript Scraper, Audio Transcriber, RSS and Apple

Podcast and Audio Transcriber is an Apify Actor that converts public podcast episodes from RSS feeds or Apple Podcasts show URLs and direct audio or video file links into plain text transcripts and optional timestamped segments without a podcast login or transcription API key.

No podcast login. No transcription API key. CPU transcription inside the Actor. Researchers and content teams can use podcast transcription, RSS transcript scraping and Apple Podcasts transcripts to turn spoken audio into structured episode rows.

Only a delivered transcript containing at least 20 characters qualifies for minute charges. Failed downloads, nonaudio responses and empty or shorter transcripts produce no row and no transcript charge.

Pricing

Prices are per event. You pay only for audio that was turned into a transcript row.

Event	Configured price	You pay when
apify-actor-start	$0.00005	The platform records its synthetic start event
transcript-minute	$0.012 per started audio minute	A qualifying transcript row has been delivered

The transcript price is $0.72 per hour of transcribed audio. Duration rounds up separately for each episode. A 280 second episode counts as five minutes and costs $0.06 in transcript events, plus the platform start event. Many short episodes can cost more than $0.72 per combined hour because each rounds up.

Local and free runs without pay per event pricing deliver rows with charged set to false. The start event is managed by the platform and never charged by Actor code. The minute charge counts the audio processed, including silence, rather than CPU wall time or only spoken segments.

Public podcast transcription without a separate service

Supply public RSS or Apple Podcasts show URLs. Transcription uses faster-whisper on CPU inside the Actor. The default base model is downloaded when the Docker image is built, so default runs load it from disk.

{
  "podcasts": ["https://feeds.npr.org/500005/podcast.xml"],
  "maxEpisodes": 1,
  "maxMinutesPerEpisode": 10,
  "model": "base",
  "includeSegments": true
}

tiny favors speed and small offers a larger model with greater CPU cost. Those optional models download on first use and may add startup time.

Audio and video file links

Supply up to twenty public HTTP or HTTPS file links in audioUrls for audio to text. Supported formats include mp3, m4a, wav, ogg, flac, mp4, webm and mov. Video files contribute their audio track. Each file is one item with the same duration cap and delivery rules as a podcast episode. Podcasts are processed first when both inputs are supplied.

{
  "audioUrls": ["https://example.com/interview.mp4"],
  "maxMinutesPerEpisode": 10,
  "includeSegments": true
}

File rows have sourceType set to file, empty podcastTitle and feedUrl, and null publishedAt. The URL path supplies episodeTitle and the full URL supplies episodeGuid and audioUrl.

How it works

1. Resolve Apple show IDs through the public iTunes lookup endpoint, or read the supplied RSS feed directly.
2. Select audio enclosures newest first, up to maxEpisodes for each podcast.
3. Follow media redirects with a browser User-Agent and accept audio, video, generic binary or missing content types. Reject text and JSON responses. Downloads are capped at 200 MB.
4. Use ffprobe to measure audio and ffmpeg to convert only the first maxMinutesPerEpisode minutes to 16 kHz mono WAV.
5. Transcribe with four CPU threads, int8 computation, beam size one and voice activity detection.
6. Deliver a qualifying row first, then charge the rounded-up audio minutes. Stop cleanly when the remaining charge allowance cannot cover an episode.

No new episode starts within 90 seconds of the run timeout. Before transcription the charge allowance is checked against the prepared audio duration. If an episode cannot fit the allowance, the run stops instead of processing only an affordable part.

Apple episode links with ?i= select only the newest RSS episode. Apple episode IDs are not reliably matched to RSS GUIDs or titles, so this Actor does not promise the linked episode.

What data does it extract?

Each episode or file row contains sourceType, podcastTitle, feedUrl, episodeTitle, episodeGuid, publishedAt, audioUrl, audioDurationSec, transcribedSec, truncated, language, languageProbability, model, transcript, wordCount, scrapedAt and charged. Podcast rows have sourceType set to podcast.

segments contains start, end and text when includeSegments is true. Times are seconds from the beginning of the episode. Publication dates are null when missing. audioDurationSec uses feed metadata when present and ffprobe otherwise. transcribedSec measures the prepared audio actually sent to Whisper. truncated uses the measured source duration.

charged identifies a row submitted for paid minute charging. Dataset delivery precedes charging, so an exceptional charge failure can leave a delivered row marked true. The run logs the failure and stops. The platform charge record is authoritative.

What is NOT returned

• Speaker names, speaker identification or separate speaker tracks.
• Private feeds, login-protected episodes or anything behind a paywall.
• Video frames or downloaded media files.
• Audio beyond the first maxMinutesPerEpisode minutes.
• A guaranteed transcript of the specific episode referenced by an Apple episode ID.
• Human-verified text, perfect punctuation or guaranteed spelling of names.

Start a run

1. Open the Actor and keep the prefilled NPR feed, one episode and a ten-minute cap.
2. Select a language or leave it empty for automatic detection.
3. Start the run and read the default dataset when it finishes.
4. Export rows as JSON, CSV or Excel.

The prefilled input selects one episode. The general defaults select up to three episodes per podcast and cap each at 60 minutes. The default run allocates 4096 MB and a 3600-second timeout.

Input

{
  "podcasts": ["https://podcasts.apple.com/us/podcast/npr-news-now/id121493675"],
  "maxEpisodes": 3,
  "maxMinutesPerEpisode": 60,
  "model": "base",
  "language": "en",
  "includeSegments": true
}

Field	Type	Default	Meaning
podcasts	string array	Optional, prefilled NPR feed	Zero to ten public RSS or Apple Podcasts URLs
audioUrls	string array	Optional, empty	Zero to twenty public audio or video file URLs
maxEpisodes	integer	3, prefill 1	Newest episodes per podcast, maximum 50
maxMinutesPerEpisode	integer	60, prefill 10	First N minutes only, maximum 240
model	enum	base	tiny, base or small
language	string	Empty	Lowercase ISO 639-1 code, or automatic detection
includeSegments	boolean	true	Include timestamped segments

At least one of podcasts or audioUrls must be non empty. maxEpisodes applies only to podcasts. maxMinutesPerEpisode applies to each episode or file.

Output

This complete row comes from an offline test using the supplied NPR RSS metadata and sample audio fixture with a one-minute cap. It is an actual local transcription with free-run delivery. The sample audio is longer than the feed duration, so the row preserves feed metadata while measuring the transcription cap independently.

{
  "sourceType": "podcast",
  "podcastTitle": "NPR News Now",
  "feedUrl": "https://feeds.npr.org/500005/podcast.xml",
  "episodeTitle": "2am ET, October 06 | The latest headlines from NPR",
  "episodeGuid": "9fdf9910-5b0a-499c-b5ae-780ff78cbb0e",
  "publishedAt": "2026-10-06T06:09:50+00:00",
  "audioUrl": "https://prfx.byspotify.com/e/play.podtrac.com/npr-500005/npr.simplecastaudio.com/368f8510-314e-46f1-8d26-5b87ed6ab6eb/episodes/4c6fe20f-5d14-4a4b-b807-2bb26505b888/audio/128/default.mp3?awCollectionId=368f8510-314e-46f1-8d26-5b87ed6ab6eb&awEpisodeId=4c6fe20f-5d14-4a4b-b807-2bb26505b888&feed=O9WlY7a5&t=podcast&e=nx-s1-20261006-0200-long&p=500005&d=280&size=4585474",
  "audioDurationSec": 280.0,
  "transcript": "This is NPR News Now. The latest headlines every hour. Live from NPR News, I'm child Snyder. President Trump is back at the White House after he signed an executive order named at lowering diesel fuel costs. Trump signed the order during a Monday evening campaign rally in Nebraska as NPR's Franco Ordonia's reports. President Trump told the audience in the middle of agriculture country that he was signing the order to help farmers. The rising cost of fuel has been a significant challenge for Republicans heading into the midterms. Prices for diesel used by trucks have been hit particularly hard due to the wars in both Iran and Ukraine. The president did not share details but the order is expected to expand access to diesel, including so-called red dye diesel fuel which is used by the agriculture industry and not subject to the same state and federal highway taxes as regular diesel. A number of states, including Texas.",
  "segments": [
    {
      "start": 0.0,
      "end": 4.4,
      "text": "This is NPR News Now."
    },
    {
      "start": 4.4,
      "end": 7.28,
      "text": "The latest headlines every hour."
    },
    {
      "start": 7.28,
      "end": 11.08,
      "text": "Live from NPR News, I'm child Snyder."
    },
    {
      "start": 11.08,
      "end": 15.8,
      "text": "President Trump is back at the White House after he signed an executive order named at"
    },
    {
      "start": 15.8,
      "end": 18.24,
      "text": "lowering diesel fuel costs."
    },
    {
      "start": 18.24,
      "end": 23.36,
      "text": "Trump signed the order during a Monday evening campaign rally in Nebraska as NPR's Franco"
    },
    {
      "start": 23.36,
      "end": 24.84,
      "text": "Ordonia's reports."
    },
    {
      "start": 24.84,
      "end": 28.8,
      "text": "President Trump told the audience in the middle of agriculture country that he was signing the"
    },
    {
      "start": 28.8,
      "end": 30.64,
      "text": "order to help farmers."
    },
    {
      "start": 30.64,
      "end": 36.32,
      "text": "The rising cost of fuel has been a significant challenge for Republicans heading into the midterms."
    },
    {
      "start": 36.32,
      "end": 40.88,
      "text": "Prices for diesel used by trucks have been hit particularly hard due to the wars in"
    },
    {
      "start": 40.88,
      "end": 43.2,
      "text": "both Iran and Ukraine."
    },
    {
      "start": 43.2,
      "end": 48.44,
      "text": "The president did not share details but the order is expected to expand access to diesel,"
    },
    {
      "start": 48.44,
      "end": 53.76,
      "text": "including so-called red dye diesel fuel which is used by the agriculture industry and not"
    },
    {
      "start": 53.76,
      "end": 58.12,
      "text": "subject to the same state and federal highway taxes as regular diesel."
    },
    {
      "start": 58.12,
      "end": 60.0,
      "text": "A number of states, including Texas."
    }
  ],
  "language": "en",
  "languageProbability": 0.9997122883796692,
  "transcribedSec": 60.0,
  "truncated": true,
  "model": "base",
  "wordCount": 159,
  "scrapedAt": "2026-10-06T06:50:00.667370+00:00",
  "charged": false
}

Use from Claude, ChatGPT and any MCP agent

Add it through the Apify MCP server and ask your agent to transcribe the newest episodes from a public feed.

https://mcp.apify.com/?tools=george.the.developer/podcast-transcript-scraper

From Python with an Apify token for platform access.

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("george.the.developer/podcast-transcript-scraper").call(
    run_input={
        "podcasts": ["https://feeds.npr.org/500005/podcast.xml"],
        "maxEpisodes": 1,
        "maxMinutesPerEpisode": 10,
    }
)
rows = client.dataset(run.default_dataset_id).list_items().items
print(rows)

With curl, wait for the run and return the dataset rows.

curl -X POST "https://api.apify.com/v2/acts/george.the.developer~podcast-transcript-scraper/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"podcasts":["https://feeds.npr.org/500005/podcast.xml"],"maxEpisodes":1,"maxMinutesPerEpisode":10}'

An Apify token authenticates platform access. No Apple, podcast host or external transcription credential is needed.

Integrations

Clay. Use https://api.apify.com/v2/datasets/DATASET_ID/items?format=json as an HTTP enrichment source and map episodeTitle, transcript, language and feedUrl to columns.

n8n and Make. Select the Actor in an Apify node or module, enter feed URLs and read the default dataset in the next step.

Google Sheets. Export rows as CSV or use the Apify Google Sheets integration. JSON preserves nested timestamped segments.

Use cases

• Researchers collect transcripts from selected public feeds and retain episode GUIDs and source links.
• Content teams review timestamped passages before preparing episode summaries.
• Search builders index plain text with podcast titles and publication dates.
• AI agents retrieve transcript rows through MCP and cite the original episode metadata.

FAQ

How accurate is transcription?

Whisper base on CPU produces automated text and English works best. Names, accents, background noise and overlapping speech can cause errors. Language probability estimates language detection, not transcript correctness. Review passages before quoting them.

Does an Apple episode link select that exact episode?

No. A link with ?i= selects the newest RSS episode only. Use a show URL for several newest episodes. The Apple episode identifier is not mapped to an RSS GUID.

Will the whole episode be transcribed?

Only when it fits maxMinutesPerEpisode. Longer audio is trimmed before transcription and marked truncated true. The duration cap applies to each episode separately.

What happens when a download or transcription fails?

The episode is skipped with a log line and no transcript minute charge. A nonaudio response or fewer than 20 transcript characters produces no dataset row. A failed dataset push never triggers a minute charge.

How fast is it?

On Apify with the base model and 4 GB of memory, a 5.3 minute news episode finished in 75 to 97 seconds per run on October 6, 2026, including download, model start and transcription. Longer episodes take longer in proportion. Downloads, CPU contention and model choice affect total time.

Is an API key needed?

No transcription or podcast API key is needed. Apify API and MCP access require platform authentication. The default model is bundled in the built image. Optional models need access to download their weights.

Can I get speaker labels?

No. The Actor does not return speaker names, speaker identification or separate speaker tracks. Each row has the plain transcript and, when includeSegments is true, timestamped segments with start, end and text.

Can an AI agent call this actor?

Yes. Add it through the Apify MCP server and ask your agent to transcribe the newest episodes from a public feed. Podcast rows carry the episode title, feed URL, publish date and transcript, so the agent can cite the source episode. It also runs from Python with the Apify client or from curl through the Apify API.