These look like the same product and are not. Both take a podcast link and hand back text through an API an agent can call. The difference is where the text comes from: TranscriptFetch resolves the link to the episode audio and transcribes it, so you get a fresh machine transcription. Spoken returns the transcript that already exists for that episode, and resolves the voices to real names, "Andrew Huberman" rather than "Speaker 1". They fail in different places, and the right one depends on whether you need coverage or fidelity.
Price
TranscriptFetch is cheaper per transcript than Spoken, and not by a little. Their plans start at $5/month and their per-transcript rate sits well below anything on Spoken's ladder. If cost per transcript is your deciding variable, they win.
Two things to weigh against the sticker price:
- Their credits are subscription-bound. TranscriptFetch sells monthly plans. Spoken sells one-time packs, and the credits never expire.
- Re-fetching costs nothing on Spoken. Pulling an episode again costs 0 credits. If your pipeline reprocesses its corpus for a new chunker or embedding model, that is the difference between a one-time cost and a recurring one.
Side-by-side comparison
| TranscriptFetch | Spoken | |
|---|---|---|
| Transcript source | Generated from the audio | The transcript published with the episode |
| Speaker names | No speaker identification | Real names, resolved per turn |
| Coverage | Anything with audio behind the link | Episodes that have a published transcript |
| Billing model | Monthly subscription | One-time credit packs |
| Credit expiry | Tied to the subscription period | Never expire |
| Re-fetching an episode | Costs a credit each time | Free, always |
| Output format | JSON segments | Markdown, speakers bold, timestamps per turn |
| Sizing an archive first | Not offered as a separate step | Free catalogue listing before you spend |
Where TranscriptFetch wins
If an episode has audio, they can return text for it, including shows that publish no transcript at all. Spoken returns a 404 for those.
If you need every episode of an arbitrary list of shows, Spoken cannot promise that and they can. They also accept a wider range of input links, and at high monthly volume their unit price is hard to beat.
Where Spoken wins
Machine transcription and a published transcript are not the same artifact. The published one carries the spelling of names, products, companies and jargon as published with the episode. A speech model guesses at those, and guesses worst on the proper nouns you are most likely to search for later.
- Real speaker names. TranscriptFetch ships no speaker identification. For interview and panel shows, unattributed text is close to useless for retrieval, because "who said this" is usually the question. Resolving it yourself means a second model pass over every episode.
- Markdown, not segments. The response goes straight into a prompt or a chunker without a normalisation step.
- You can size the job before paying. Listing a show's catalogue is free and returns only the episodes that are actually fetchable, so an archive quote is knowable up front.
Pick the right tool
Pick TranscriptFetch for
- Coverage of shows that publish no transcript
- Very high monthly volume where unit cost dominates
- Mixed sources beyond podcasts
- Cases where approximate text is good enough
Pick Spoken for
- Archiving a specific show properly, once
- Anything where "who said it" matters
- Correct spelling of names, products and jargon
- RAG and search, where a wrong proper noun is unrecoverable
- Buying credits that never expire
FAQ
Is TranscriptFetch cheaper than Spoken?
Per transcript, TranscriptFetch is cheaper than Spoken: its plans start at $5/month and the effective rate per transcript is below Spoken's. What Spoken offers instead is the published transcript rather than a machine transcription, named speakers, credits that never expire, and free re-fetches.
Does TranscriptFetch identify speakers?
No. TranscriptFetch has no speaker-identification feature, so a multi-speaker episode comes back without attribution. Spoken resolves speakers to real names as part of the response.
Which one has better coverage?
TranscriptFetch has clearly broader coverage than Spoken: it transcribes the audio, so it can return something for almost any episode. Spoken only serves episodes that have a published transcript and returns a 404 otherwise, which is never charged.
Can I use both?
Yes, Spoken and TranscriptFetch together are the sensible pattern for a broad archive: Spoken for shows that publish transcripts, where you want named speakers and exact spelling, and a transcription service for the rest. They fail in different places, so the union covers more than either alone.
How do I know which episodes Spoken can actually return?
Listing a show's back catalogue on Spoken costs no credits and returns only the episodes that are fetchable, so you can size and price an archive before spending anything.
TL;DR: TranscriptFetch is cheaper and covers more, because it transcribes audio. Spoken returns the published transcript with real speaker names, credits that never expire, and free re-fetches. Pick on fidelity vs coverage, not on price.
"I used Spoken to add every My First Million episode to my knowledge base, with a cron to pull new ones. Now I can enjoy the podcast on a run, then chat with Claude about it later — every episode saved and accessible in my Claude sessions."
Thousands of transcripts fetched by developers building summarizers, RAG pipelines, and podcast tools