Transcription APIs you can pay per call

What a transcription API does, and who calls it

A transcription API turns spoken content into text. Two products answer that request with different contracts. Speech-to-text takes audio or video the caller supplies, as bytes, a URL or an upload, and runs a recognition model on it. A YouTube transcript route takes the address of a public video and returns text for it; for 5 services on this page, the endpoint description says that text is read from captions already published with the video.

An agent calls speech-to-text to read a voice note, a meeting recording or a support call. It calls a video transcript route to summarize a talk, index a channel or cite a passage. The first depends on the quality of your audio and the recognition model. The second depends on whether the video has captions and how good they are.

The table covers 11 services with a matching priced transcript endpoint as of October 26, 2026: 4 services sell speech-to-text on supplied media and 7 services return transcripts of public YouTube videos. It leaves out 4 candidates that pass the listing checks but list no matching priced transcript endpoint in the captured list.

What the 11 transcription services actually return

The classification follows each endpoint's documented product. A service with more than one product appears in more than one row, and each quote comes from the cheapest priced endpoint of that type, with its route attached.

Each price is the cheapest endpoint of that service for that kind of data: the same number wherever this page quotes it.

Timestamped segments and subtitle files describe the shape of the returned text, not a separate recognition step. A passage search inside a transcript is useful for citations but does not return the full text. Source labels report what a description states: text read from captions, or a named recognition model. They are provider statements, not accuracy measurements by x402-list.

The 11 transcription APIs and their matching endpoint prices

One row per service, ordered by measured buyers. The price is the cheapest speech-to-text or video transcript route of the service, named under the price, never an unrelated tool on the same host.

$/CALL is the price of the endpoint named under it: the cheapest speech-to-text on audio or video you supply, transcript of a public youtube video and transcription (route named by its path) endpoint of the service or, for a service without one, its cheapest transcript endpoint. Never the cheapest tool of a gateway. Other kinds of data have their own price in the tables below. TRANSCRIPT ENDPOINTS counts the transcript endpoints out of all the endpoints the service lists. BUYERS 30D counts distinct wallets that paid the whole service on-chain in 30 days, not only its transcript endpoints. "Generalist" marks a gateway that sells more than three unrelated kinds of tools: it is never called the cheapest.

The buyer figure counts wallets that paid the whole service, so for a gateway it does not show demand for transcription. A low monitoring latency describes our unpaid checks, not the time needed to transcribe an hour of audio.

How to choose a transcription API

Your own audio or a public video

Start from what you hold. If you have the recording, you need speech-to-text and the right to process it. If you have only a public YouTube address, a caption-based route is usually the direct path, but it returns nothing useful for a video that was never captioned. Do not compare the two prices as if they bought the same work.

Duration, size and input format

Stated limits differ by provider in both duration and file size; the service notes below give each limit a description states. Check whether the route takes base64 in the body, a public URL or a multipart upload, and which formats it accepts. A URL input needs media the provider can reach; base64 input inflates the request size.

Timestamps, subtitles and chapters

Plain text is enough for a summary. Citations, search and subtitles need timed segments, and a subtitle workflow is simpler when the route returns SRT or WebVTT directly. Check the timestamp unit and whether chapters come from the creator or from the provider.

Missing captions, silence and billing

Read what each provider says about failure cases: a video without captions, a file with no clear speech, a timeout. Where a description says whether such a request is billed, the service notes below quote it; otherwise ask the provider. A daily allowance or a free first call is a provider statement, separate from the per-call price in the table.

Language and accuracy claims

Language counts, model names and statements such as verbatim output come from provider descriptions. x402-list has not measured word error rates. Test with recordings that contain your accents, names and background noise before choosing.

Which endpoint fits which job

Start with the job, then compare the candidates whose classified endpoints match it. The price in a job list is tied to those types and can differ from the service's price in the main table.

Transcribe your own recording

Check the input route first: base64 in the body, a public URL or a multipart upload. Compare the stated maximum duration and file size with your recordings, and decide how to split longer audio. Keep the original file, the language and the model setting with the transcript so a disputed passage can be checked.

Candidates today, most buyers first: Cluster Protocol ($0.006/call, generalist, 719 buyers) · agentutility ($0.1/call, generalist, 36 buyers) · NeuroDynamic Audio ($0.01/call, 7 buyers) · Transcribe — audio to transcript, subtitles and a cut list ($0.05/call)

Get the transcript of a public YouTube video

Ask what happens when a video has no captions in the requested language, whether that request is charged and whether auto-generated captions are accepted. A caption-based route returns the text published with the video; it does not recognize speech that was never captioned.

Candidates today, most buyers first: agdata ($0.024/call, generalist, 19 buyers) · SwarmIO Cloud ($0.02/call, generalist, 10 buyers) · Marketplace for AI Agents ($0.003/call, generalist, 7 buyers) · Melchior Video Context ($0.01/call, 3 buyers) · Kenward payable API ($0.01/call, 2 buyers) · Fly Labs Agentic Market ($0.03/call, 1 buyer) · YouTube Transcript APIs for AI Agents ($0.004/call)

Cite or jump to a moment

Timestamped segments let an agent quote a passage with a link to the moment it was said. Check the unit of the timestamps, whether segments overlap and how a passage search returns its matches before building citations on them.

Candidates today, most buyers first: agdata ($0.024/call, generalist, 19 buyers) · SwarmIO Cloud ($0.02/call, generalist, 10 buyers) · Marketplace for AI Agents ($0.003/call, generalist, 7 buyers) · Melchior Video Context ($0.005/call, 3 buyers) · Kenward payable API ($0.01/call, 2 buyers) · Fly Labs Agentic Market ($0.03/call, 1 buyer) · YouTube Transcript APIs for AI Agents ($0.004/call)

Produce subtitle files

SRT and WebVTT are timed text formats with their own line and timing rules. Validate a returned file in the player that will use it, and review line breaks and reading speed before publishing subtitles to viewers.

Candidates today, most buyers first: agdata ($0.024/call, generalist, 19 buyers) · Transcribe — audio to transcript, subtitles and a cut list ($0.05/call)

Service details and limits to check

The notes below are read from each provider's description and endpoint list captured on October 22, 2026. They attribute advertised capabilities; x402-list has not tested the accuracy of the returned transcripts. Prices, uptime and buyers come from the current page model.

Cluster Protocol

Cluster Protocol is an inference gateway that lists POST /v1/audio/transcriptions as audio transcription, priced at $0.006. The same host lists speech synthesis on a separate route; that product turns text into audio and is not part of this table. The endpoint description is short: it names speech-to-text but does not state accepted formats, a maximum duration or the shape of the returned text. Read the provider's schema before sending a recording, and test with a clip representative of your audio quality. Its buyer count covers payments to the whole gateway, so it does not show how often the transcription route itself is used.

1 transcript endpoint of 9 · $0.006/call (POST /v1/audio/transcriptions) · Base · uptime 30d 100% · 264 ms · 719 buyers in 30 days

NeuroDynamic Audio

NeuroDynamic Audio lists POST /v1/audio/transcriptions for a voice note or short recording, at $0.01. Its description sets the limits at 60 seconds and 12 MiB per request, accepts WAV, MP3, FLAC or Ogg, and returns a JSON transcript; it also labels the quote introductory pricing. Split a longer recording before sending it, and keep the order of the pieces so the transcript can be joined. The host also sells narration and private voice cloning, which produce audio rather than text and are excluded from the transcription type.

1 transcript endpoint of 3 · $0.01/call (POST /v1/audio/transcriptions) · Base · uptime 30d 99.7% · 297 ms · 7 buyers in 30 days

Transcribe — audio to transcript, subtitles and a cut list

Transcribe lists POST /transcribe for an audio or video URL and describes three outputs in one response: the transcript, ready-made WebVTT subtitles and an editor's cut list that marks stretches of dead air. The captured endpoint price is $0.05. The service description and the endpoint description state different flat amounts, so treat the payment requirement of the exact request as the reference. Sources are limited to 800 KB, and the provider says a file with no clear speech is still billed because the recognition ran; its reply flags no speech or low confidence.

1 transcript endpoint of 1 · $0.05/call (POST /transcribe) · Base · uptime 30d 100% · 1307 ms · 0 buyers in 30 days

agentutility

agentutility lists POST /video-to-text, which takes a video URL, extracts the audio and transcribes it with a model it names as Whisper v3 large. The quote is $0.1. The description claims automatic detection of more than 90 languages, a translate-to-English mode and files up to 60 minutes or 500 MB. It points to a separate speaker-diarization endpoint for speaker labels, but that route is not in the endpoint list read for this page, so speaker separation is not priced here. The host is a generalist catalog with document, image and web tools.

1 transcript endpoint of 9 · $0.1/call (POST /video-to-text) · Base/SOL · uptime 30d 99.8% · 1502 ms · 36 buyers in 30 days

agdata

agdata lists GET /x402/youtube-transcript at $0.024. Its description is explicit about the source: it returns the subtitles already published for the video, with no speech-to-text, as timed JSON segments, plain text, SRT or WebVTT, together with title, channel, duration, views and publish date. The caller chooses one of ten subtitle languages or any language. The provider says a video without subtitles in the requested language is not charged. A video without published subtitles therefore has no transcript from this route; use a recognition product on the media itself where you have the rights to do so.

1 transcript endpoint of 22 · $0.024/call (GET /x402/youtube-transcript) · Base/POL/SOL · uptime 30d 100% · 57 ms · 19 buyers in 30 days

SwarmIO Cloud

SwarmIO Cloud lists POST /api/v1/video/transcript for the captions of one YouTube video as timestamped segments with title and duration, at $0.02 per call after a daily allowance the provider describes. It says a video without a usable caption track, for example private, deleted or with captions disabled, is answered before any payment request. That free allowance is the provider's statement; the price in this table is the captured per-call quote. The host also sells news, market and filing feeds, which are unrelated to transcripts.

1 transcript endpoint of 19 · $0.02/call (POST /api/v1/video/transcript) · Base · uptime 30d 99.9% · 844 ms · 10 buyers in 30 days

YouTube Transcript APIs for AI Agents

YouTube Transcript APIs for AI Agents lists three POST routes, /youtube-captions, /youtube-transcript and /youtube-video-to-text, and its YouTube transcripts start at $0.004. All three take a YouTube URL or video ID. The captions route extracts captions and subtitles, and the video-to-text route says it converts the video into clean text using the captions available; the transcript route does not say whether its text comes from captions or from recognition. The captions and transcript routes mention optional timestamps and language selection. Despite the name, video-to-text here is caption extraction, not recognition run on the audio track.

3 transcript endpoints of 3 · $0.004/call (POST /youtube-captions) · Base · uptime 30d 99.9% · 622 ms · 0 buyers in 30 days

Melchior Video Context

Melchior Video Context lists POST /v1/youtube/transcript, which turns a public YouTube video into transcript chunks with timestamp citations, at $0.01. It also lists GET /v1/youtube/search at $0.005 to find passages matching a phrase inside one video, again with timestamps. A third route that only normalizes text you already have into chunks is not a transcript product and is left out. The descriptions do not say whether the transcript is read from captions or recognized from audio.

2 transcript endpoints of 3 · $0.01/call (POST /v1/youtube/transcript) · Base/SOL · uptime 30d 100% · 190 ms · 3 buyers in 30 days

Fly Labs Agentic Market

Fly Labs Agentic Market lists POST /api/agents/transcribe at $0.03. The path name suggests recognition, but the description says it returns verbatim captions for public videos that have captions, manual or auto-generated, in any language. The response is described as full text, language, duration, word count, time-indexed paragraphs, creator-defined chapters and video metadata. Auto-generated captions are themselves machine output from the video platform, so a statement that the text is verbatim describes fidelity to those captions rather than to what was actually said.

1 transcript endpoint of 2 · $0.03/call (POST /api/agents/transcribe) · Base · uptime 30d 99.8% · 661 ms · 1 buyer in 30 days

What the calls cost at different volumes

The median selected endpoint costs $0.01 per call. At that same request shape, 100,000 calls would cost $1,000.00. This is a comparison aid, not an invoice: a long recording split into pieces is several calls, and a route priced per request costs the same for a short clip and for a file at its limit.

Each service at the price of its main transcript endpoint, the one named in the table above.

Budget for the failure cases as well. A request that returns no speech, a retry after a timeout or a second call for timestamps can each add a charge, depending on the provider's terms. Keep the route, the duration of the media and the price of each call in your job record.

Inspect a request and its payment evidence

For a short recording, NeuroDynamic Audio documents POST /v1/audio/transcriptions, with a speech-to-text quote of $0.01. The description sets a 60-second and 12 MiB limit and lists WAV, MP3, FLAC and Ogg; read the provider's schema for the body fields before sending audio.

The listed method and description establish the advertised contract. This explanation does not reproduce an HTTP 402 challenge or certify a paid response. Open the service page's payment evidence and request the challenge for the exact body before authorizing payment. Compare resource, method, network, asset and amount together, and do not reuse a challenge obtained for a different route on the same host.

What this comparison measures

The listing requires an online, payment-ready service, a usable description, an advertised price and at least 90% measured uptime over 30 days. Task selection then reads the endpoint list and keeps documented transcript routes. A podcast search, an HTTP receipt that stores the transcript of a page fetch, a text chunker or a route whose listing repeats only its own path does not count as a transcript endpoint, and a transcript product named only in a service description is not priced without a matching route.

Uptime and average latency come from monitoring checks. Buyer counts are distinct paying wallets observed for the whole service in the window. A successful unpaid HTTP 402 response shows a payment requirement, not a correct transcript after payment. Test a representative file or video and define what counts as a usable result before putting a recurring workflow on any endpoint.

Questions

How to pay

Each service answers an unpaid call with HTTP 402 and its price; your agent signs a USDC payment on the listed network and retries. No API key, no account, no subscription. See what x402 is.