Transcription services for under 50 cents an audio hour

Find the answer to what you searched

This page answers the buying questions in one place: which service is best for you, how to spend the least, transcribing online on any device and outsourcing a steady stream of recordings. The pages below go deeper on one kind of recording or one price question.

What comes back

A transcript with a timestamp on every line, exported as TXT, DOCX, SRT, VTT or JSON, plus a summary, key quotes and chapters. Speakers are identified and named in the summary, key quotes and chapters. The key quotes show each person's words without reading the whole transcript; click any timestamp to check a quote against the audio.

What it costs

You buy minutes once and spend them on recordings. Each transcript uses the length of its recording, and unused minutes never expire, so an uneven month costs the same as a steady one. There is no seat fee and no subscription unless you want the Workspace plan.

Workspace is $19 a month for people and teams who transcribe steadily. Every recording starts with a free preview of its opening minutes before you pay; a pack from $4 transcribes the rest. Every price is on the pricing page.

For scale: a 40-hour interview project fits in the $24 pack with minutes left over. At Rev's human rate of $1.99 a minute (checked 25 October 2026), the same 40 hours costs $4,776.

Cheap transcription: how to spend the least

The cheapest hour of transcription is the one you do not pay twice for. Four habits keep the bill at the bottom of the range above:

1. Read the free preview first. Every recording starts with a free preview of its opening minutes, so you see that the audio was understood before you spend anything.
2. Buy the pack that matches the year, not the week. Minutes never expire, so the $12 or $24 pack lowers the hourly rate (about $0.36 or $0.29) without the risk of losing what you do not use this month.
3. Paste links instead of downloading. YouTube, podcast, RSS and Google Drive links go straight in; there is nothing to convert first.
4. Move to Workspace only when you transcribe every day. At $19 a month it costs less per hour than the $24 pack only once you transcribe more than about 66 hours a month; below that, packs are cheaper.

Worked out on real packs: one hour of audio this month fits the $4 pack with more than seven hours left; ten one-hour interviews (600 minutes) fit the $12 pack; a 40-hour project (2,400 minutes) fits the $24 pack. There is no seat fee, so one balance covers everyone who uploads.

Turnaround

Processing takes about 2 minutes per hour of audio, and most files finish in under a minute. There is no rush fee because there is no slow lane to pay your way out of. A ten-hour recording is one job, not ten. Busy hours add queue time, and the tracker on the upload page shows it while you wait.

Batches. Upload files or paste links one after another; each becomes its own transcript, drawing on the same credit balance. A file can be up to 10 hours or 5 GB. The box on this page takes files up to 500 MB; larger files continue on the upload page. For a backlog, see bulk transcription.

What you get

• The transcript, with timestamps, in a library you can search. Click a line to hear that moment.
• Exports as TXT, SRT, VTT, DOCX and JSON: DOCX to edit, SRT or VTT for captions, TXT for search and analysis tools, JSON with timings for your own software.
• A summary of the recording: the thesis and takeaways, chapters with timestamps, the names, numbers and dates that were said, the questions it answers, and a chat over the transcript.
• 99+ languages, detected automatically, so a Spanish interview or a French lecture needs no setting.
• Any source: audio or video files (MP3, MP4, M4A, WAV, FLAC, OGG, WEBM, MOV and more), a YouTube, podcast, RSS, Google Drive, Twitch or Kick link, or a microphone recording.
• Privacy. Recordings are transcribed on our own servers and are never used for training. You can delete a recording and its transcript at any time.

What it is not: the transcript is one continuous timestamped text and does not mark which person is speaking, and it is not a live caption feed for a stream that is still running. If you need either, the comparison below shows who sells it.

Online, on any device

Online means the transcription runs on our servers and the text comes back in a browser tab. There is nothing to download and no fast computer to own; Windows, a Mac, a Chromebook, an iPhone or an Android phone all work the same. The page that takes the recording shows the transcript, searches it, plays any line and downloads it.

On a phone. Most recordings already live on a phone, so there is no need to move them to a computer first.

• iPhone: open this page in Safari, tap the box and pick the recording from Files or Voice Memos (share the memo to Files first if it is not listed). Step by step: Voice Memos to text.
• Android: open the page in Chrome, tap the box and pick the file from your recorder app's folder or Downloads. Step by step: Android recorder to text.
• A link instead of a file: copy the share link from YouTube or a podcast app and paste it. Nothing is downloaded to the phone, which saves data on long videos.

Browser or desktop app. Desktop transcription apps run a speech model on your own computer. They are free or cheap, work offline and nothing leaves the machine, but they need installing, disk space and a computer powerful enough for the larger models. Pick one if your recordings must never leave your computer or you transcribe without a connection. Pick an online service if you are on a phone or a work laptop where you cannot install software, your computer is slow, or the recording is a link rather than a file.

Privacy and limits. Recordings are transcribed on our own servers; audio is never sent to a third-party AI service and never used to train any model. Delete a recording and its transcript at any time, and deleting your account removes every record tied to your email. One file can be up to 10 hours or 5 GB, in MP3, MP4, M4A, WAV, FLAC, OGG, WEBM, MOV and more; for very long material see transcribing long audio.

See a finished transcript first

These are real transcripts of public recordings, open to anyone, so you can judge the output before you spend anything.

• Pride and Prejudice, a public-domain LibriVox audiobook reading in English.
• Candide, a French audiobook, to see a language other than English.
• LINUX Unplugged 661, a technical podcast with several hosts and plenty of jargon.
• Inside SQLite and TPU origins and the future of AI, write-ups of public talks with chapters and quotes.
• More than a thousand more on the public recordings index.

Which transcription service is best for you

There is no single best service, only a best one for each job, including jobs we do not do.

• Recordings you already have, at the lowest price per hour: WhipScribe. One-time packs from $4 for 500 minutes, about $0.29 an audio hour on the $24 pack, 99+ languages and files up to 10 hours.
• A person must check every word: a human service such as Rev. Rev lists human transcription at $1.99 a minute (checked 25 October 2026). Pay it for court filings, certified records and very poor audio.
• Notes from live calls: Otter. Per Otter's site it is a meeting assistant that joins calls; WhipScribe transcribes the recording afterwards, with no bot in the room.
• Editing a video or podcast by editing its text: Descript. Per Descript's site it is an audio and video editor with transcription inside it. If all you need is the text, you are paying for an editor.

What decides it. Compare price per audio hour, not per month: a plan is cheap only if you use every minute it includes, while a pack costs the same per hour whether you transcribe one file this year or a hundred this week. Then ask whether a machine transcript you check is enough or a human must stand behind it, whether the audio is a meeting happening now or a recording that already exists, and whether your files are long or not in English.

WhipScribe compared with Rev, Otter and Descript

All four turn speech into text, but they are built for different jobs. Rev and Otter prices below are copied from each company's own pricing page, checked 25 October 2026; check the page before you buy, because plans change.

The fair reading. If you transcribe meetings every week, Otter's annual Pro plan ($8.33 a month for 1,200 minutes, about $0.42 an hour if you use every minute) is close to our per-hour price and is designed for that job. If you edit podcasts or videos, Descript gives you an editor, not just a transcript. If a person must check every word, Rev sells that and we do not. WhipScribe is the better buy when your volume is uneven, your recordings are long, or you want to pay a few dollars once instead of a monthly plan.

Sources, checked 25 October 2026: rev.com/pricing, otter.ai/pricing. Descript is described from descript.com without prices. Names belong to their owners; this page is not affiliated with them. More detail on the Rev, Otter and Descript comparison pages, and for other tools, each with its prices read on its own pricing page on 26 October 2026: Trint, Riverside, Notta, Brass Transcripts, Amberscript, Maestra, Verbit, tl;dv and Dragon.

AI, human or do it yourself

The three ways to turn recordings into text differ by a factor of a hundred in price. Human rates are from our human transcription directory (checked 2026-04-23).

When to choose a human service instead

WhipScribe is machine transcription, and we would rather send you elsewhere than sell you the wrong thing. Pay a human service when:

• the transcript will be filed with a court or must be certified or sworn;
• you need strict verbatim, with every false start, filler word and pause marked;
• the audio is so poor, distant or overlapping that you would re-listen to most of it anyway;
• a contract requires named transcribers, a certification, or handling in a particular country.

Our human transcription directory lists vendors with their published rates and specialties. For most other work the better value is AI first: transcribe everything, then check the passages you will quote against the audio using the timestamps, and send only those minutes to a human if you still need to.

Outsourcing a steady stream of recordings

Outsourcing transcription usually means a vendor contract: you send interviews, calls, meetings or dictation, typists work through them, and you pay by the audio minute with a turnaround in hours or business days. WhipScribe is the other way to send the work out: no quote, no minimum volume and no contract. Your team uploads or links recordings as they arrive, and transcripts land in one searchable library minutes later.

1. Send the recordings. Upload files or paste podcast, RSS, YouTube or Google Drive links, one after another; each becomes its own transcript. One file can be up to 10 hours or 5 GB.
2. Read the free preview. The opening minutes of each recording are free, so you can check it is the right file before spending credit.
3. Open the full transcripts. A pack or Workspace opens the full transcript, with a timestamp on every line, in about 2 minutes per hour of audio.
4. Export or automate. Download TXT, DOCX, SRT, VTT or JSON, or connect your own systems through the API, which comes with the $24 pack (5,000 minutes) or Workspace.

Volume pricing without a contract. For uneven volume, buy the $24 pack of 5,000 minutes (about $0.29 an audio hour): credits never expire, so a quiet month costs nothing and a busy one draws down what you already bought. For steady volume above about 66 hours a month, Workspace at $19 a month costs less per hour. Neither has a seat fee or a minimum job size, and the $4 pack covers 500 minutes, enough to run a week of real recordings through first.

The same month of audio, costed. Ten hours a month (600 minutes) fits one $12 pack, with 1,400 minutes left over; at Rev's human rate of $1.99 a minute it is $1,194. Forty hours (2,400 minutes) fits one $24 pack; at the same human rate it is $4,776. Human vendors in our directory publish $0.79–$2.50 per audio minute. Many teams split the work: everything goes through WhipScribe first, and only the few recordings a person must check go on to a vendor, so the vendor bill shrinks to the files that need it.

How to choose a transcription service

1. What is the transcript for? Notes, search, captions, analysis and drafts suit AI. A certified record needs a human.
2. How clear is the audio? AI does well on clear recordings. Names, jargon, heavy accents and crosstalk are where to check.
3. How much, and how often? Steady weekly meetings suit a subscription. Bursts of interviews, a semester of lectures or an archive suit one-time packs.
4. How long are the files? Check per-meeting and per-file limits before you buy; here one file can be up to 10 hours.
5. Which formats? DOCX to edit, SRT or VTT for captions, TXT or JSON for your own tools.
6. Where does the audio go? Ask whether recordings reach third-party AI services or train models. Here: no and no.

Transcription services by use

Related: Audio transcription services · Video transcription services · Transcription rates · Interview transcription services · Academic transcription services · Podcast transcription services · Legal transcription services · Human transcription directory · Bulk transcription · Pricing