Podcast Transcript to AI Dataset: RSS to Training Corpus
Pricing
$3.00 / 1,000 episode transcript processeds
Podcast Transcript to AI Dataset: RSS to Training Corpus
Feed in a podcast RSS feed. Get back every episode's publisher-hosted transcript as clean, structured JSON for AI training, RAG, and content intelligence. No Whisper, no audio processing.
Pricing
$3.00 / 1,000 episode transcript processeds
Rating
0.0
(0)
Developer
Moses Ndambuki
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
a day ago
Last modified
Categories
Features
Share
Feed in a podcast RSS feed. Get back every episode's publisher-hosted transcript as clean JSON for AI training, RAG, and content intelligence. No Whisper, no audio processing, no $0.006/minute transcription bills.
Paste one RSS URL. Get a full-show transcript dataset. Pay only per episode with a usable transcript.
Who this is for
- AI / ML engineers building voice, dialogue, or domain-expert corpora
- RAG developers indexing a podcast's full archive for a chatbot
- Media analysts and researchers tracking topic frequency across a show's run
- Content teams repurposing podcast content as search-indexable text
- Dataset curators assembling spoken-language training material cheaply
If you were about to pay $0.006/minute to transcribe a 500-episode archive, this costs 95% less when publishers already ship transcripts.
What you get per episode
{"podcastTitle": "The Lex Fridman Podcast","podcastDescription": "Conversations about science, technology, history...","language": "en-us","episodeTitle": "Episode 400: Elon Musk","episodeUrl": "https://lexfridman.com/elon-musk/","guid": "ep-400","publishedAt": "2026-10-15T14:00:00Z","duration": "02:15:30","audioUrl": "https://media.simplecast.com/ep-400.mp3","transcriptSource": "https://lexfridman.com/transcripts/ep-400.vtt","wordCount": 24180,"text": "Elon Musk: The most important thing I can say about...","scrapedAt": "2026-11-08T12:00:00Z"}
The output JSON shape is compatible with AI Training Dataset Builder and Sitemap to AI Corpus — chain the three freely for a mixed text + audio corpus.
How it works
flowchart LRA[Input: RSS feed URL] --> B[Fetch and parse RSS]B --> C[For each episode]C --> D{podcast:transcript tag?}D -- yes --> E[Fetch VTT / SRT / JSON / HTML]D -- no --> F[Scrape episode page for inline transcript]E --> G[Normalize to plain text]F --> GG --> H{wordCount above threshold?}H -- yes --> I[Push episode + charge per episode]H -- no --> J[Skip]
No audio transcription. This actor only extracts transcripts the podcast publisher already hosts (via the Podcasting 2.0 podcast:transcript tag, or linked from the episode page). If you need Whisper-style audio-to-text, use a separate transcription service — this one stays cheap because it never touches audio.
Quick start
Run from the Apify Console
- Click Try for free.
- Paste one or more podcast RSS URLs.
- (Optional) Set
publishedAfterto limit to recent episodes. - Click Start.
- Download as JSON, CSV, or Excel.
Run from the API
curl -X POST "https://api.apify.com/v2/acts/Turboextract~podcast-transcript-dataset/runs?token=YOUR_TOKEN" \-H "Content-Type: application/json" \-d '{"rssUrls": ["https://podnews.net/rss"],"maxEpisodesPerFeed": 50,"publishedAfter": "2024-01-01","minWordCount": 500}'
Run from Python
from apify_client import ApifyClientclient = ApifyClient("YOUR_TOKEN")run = client.actor("Turboextract/podcast-transcript-dataset").call(run_input={"rssUrls": ["https://podnews.net/rss"],"maxEpisodesPerFeed": 100,})for ep in client.dataset(run["defaultDatasetId"]).iterate_items():print(ep["episodeTitle"], ep["wordCount"])
Input fields
| Field | Type | Default | Description |
|---|---|---|---|
rssUrls | array | required | Podcast RSS feed URLs |
maxEpisodesPerFeed | integer | 100 | Cap per feed |
publishedAfter | string | — | ISO date; skip older episodes |
minWordCount | integer | 200 | Skip thin transcripts |
Pricing
Pay per episode with a usable transcript. No subscriptions.
| Volume | Price per episode | Total |
|---|---|---|
| Per episode | $0.003 | 1,000 episodes = $3 |
| 10,000 episodes | $0.003 | $30 |
Episodes without transcripts are silently skipped — you are not charged.
How it compares
| Tool | Pricing model | 1,000 episodes |
|---|---|---|
| Podcast Transcript to AI Dataset | $0.003 per episode | $3 |
| OpenAI Whisper API | $0.006 per audio minute | ~$360 (avg 60 min episodes) |
| AssemblyAI | $0.37 per audio hour | ~$370 |
| Deepgram | $0.26 per audio hour | ~$260 |
| Rev.com (human) | $1.50 per minute | $90,000 |
Why so much cheaper? This actor does not transcribe audio. It only harvests transcripts the publisher already ships via the Podcasting 2.0 standard or episode show-notes pages.
Common use cases
- AI training corpora for conversational / interview / domain-expert models
- RAG knowledge bases over a podcast network's full back catalog
- Topic modeling across years of episodes to spot trend shifts
- Searchable archives for shows that only publish MP3s + bare show notes
- Research corpora on linguistics, interviewing technique, discourse analysis
Tips for best results
- Podcasts on Podcasting 2.0 hosts (Transistor, Simplecast, Captivate, Fountain, Castopod) have the best transcript coverage
- Set
minWordCount: 500to drop one-liners and ad-only episodes - Use
publishedAfterfor incremental refreshes — pair with Apify's weekly scheduler - If a podcast has no transcripts, you'll see 0 episodes processed and 0 charges — safe to try any feed
Pairs well with
- AI Training Dataset Builder — same JSON shape; combine article corpora with podcast corpora
- Sitemap to AI Corpus — pair show-website pages with episode transcripts
- Research Citation Explorer — different buyer, same account
(Links updated as related actors ship.)
FAQ
What percentage of podcasts have transcripts?
Growing fast. ~30% of top-100 shows now ship podcast:transcript tags; most major hosts (Simplecast, Transistor, Buzzsprout, Captivate, Fountain) add them automatically. Many indie shows publish full transcripts in show notes.
Will this transcribe audio with Whisper? No. By design. Audio transcription is a different cost profile (~$0.006/min vs $0.003 per already-published episode). If you need that, use a dedicated transcription service and combine downstream.
What formats are supported? WebVTT, SRT, JSON (Podcasting 2.0 segments), plain text, and HTML transcript pages. Format is auto-detected from MIME type.
What about paywalled transcripts (Patreon, Substack private feeds)? This actor doesn't authenticate. Public RSS feeds only.
Can I run on a schedule?
Yes. Apify's scheduler + publishedAfter set to "last week" is perfect for incremental ingestion.
What if an episode has multiple transcripts (multi-language)? We prefer JSON > VTT > SRT > HTML > plain text, in that order. Language-specific selection is on the roadmap.
Does the output chain with other actors?
Yes — text, wordCount, publishedAt, language fields match AI Training Dataset Builder exactly.
Support
Open an issue on the actor's Apify page or message the maintainer. Bug reports with the failing RSS URL get the fastest turnaround.
- Source examples and docs: https://github.com/andersonjames38phd-cpu/podcast-transcript-dataset
- Community Discord (updates, support, feature requests): https://discord.gg/8SrXSCpvgj
Built and maintained by Turboextract on the Apify platform.