Roadsigns meanings

Vook.ai includes fewer editor features than GoTranscript, and it doesn’t offer any human-powered transcription or editing options. It’s also less competent at transcribing crosstalk than GoTranscript, which is important for recordings with multiple speakers.

Transcripts make searching through recordings as easy as searching any other document, simplifying finding specific quotes and action items from a meeting or jumping to the correct timestamp in audio. They help journalists, legal and medical professionals, and others who rely on interviews and spoken conversations, speeding up their work and increasing note accuracy. And they make podcasts and videos accessible to a wider audience. As a side benefit, they help listeners search through shows and even aid search engines in indexing your content.

Nearly every new tech hardware and software update today promises new AI features. With OpenAI Whisper’s open-source model already providing more-accurate transcriptions than most human transcription services offered a few years ago, expect to see future models offer even greater accuracy.

For anyone with disabilities or mobility challenges, smart devices can be profoundly empowering and can enable more independent living.

It takes, on average, four minutes to transcribe one minute of audio by hand, according to Rev, other transcription services, and anecdotes shared by transcription professionals. That time commitment paired with the training and experience needed to transcribe accurately makes human-transcribed audio typically cost $1 to $3 per minute of audio. Human transcription gives you the highest-quality results, with experience-honed accuracy and the human intuition required to guess the correct homonym or a less clearly pronounced name.

Transcription is now a commodity—a default feature in the work-centered apps so many of us use on a daily basis. Zoom, Google Meet, and Microsoft Teams transcribe meetings. Slack adds transcripts to audio shared in chat. Even YouTube auto-captions videos.

AI has raised the bar for transcription—a notion borne out by the fact that even human transcription is, on average, more accurate today than it was a half-decade ago. Steve Jobs liked to call computers “a bicycle for our minds,” a metaphor I couldn’t shake while testing these services. Today’s human transcription services often use AI for a first-pass before humans clean up the copy, showing one way that humans and AI can work best together.

Paid transcription services should offer transcripts in a variety of formats, including Word and standard caption formats.

Wirecutter is the product recommendation service from The New York Times. Our journalists combine independent research with (occasionally) over-the-top testing so you can make quick and confident buying decisions. Whether it’s finding great products or discovering helpful advice, we’ll help you get it right (the first time).

It was 98.4% accurate in our tests, but it didn’t handle speech impediments or crosstalk well—especially compared with our top pick. It also didn’t clean up filler speech or identify speakers by default, though it does include AI features to add those after the core transcription has been completed.

Happy Scribe offers both human and AI transcription, and the accuracy of its human transcription came in just under GoTranscript’s AI accuracy. It handled crosstalk well, and text was highly readable, but it was often altered from the spoken record. “Did they set a timeline?” was changed to “What was the timeline that they set?” That’s close, but the meaning was changed. Its transcript editor app looks more modern than our top pick’s and includes collaboration features it doesn’t (including comments). However, multiple people cannot edit a transcript at the same time (as you could in Google Docs, say).

It’s fast. In our tests, GoTranscript AI took 6 minutes to transcribe our entire batch of recordings —less than one minute per minute of audio, which was faster than Reduct and Rev AI, though slower than Vook.ai, OpenAI Whisper, and Descript. But GoTranscript’s human transcripts came back between 58 minutes and 2.5 hours after we placed our order—significantly faster than the anticipated one-day turnaround, similar to Rev’s human transcription, and faster than Transcription Panda (which took 6 hours 30 minutes for similar transcripts with their one-day turnaround service).

It doesn’t include speaker tracking or timestamps, which may also make it less attractive for some use cases. But it’s the transcription model to beat today, and we’d be remiss not to mention it.

AmericanRoad Sign

GoTranscript includes a basic online editor where you can download the transcript in plain text, PDF, or subtitle formats, customize speaker names, and edit the text. You can listen to your audio synced with the text, though you may need to re-upload your audio if it’s been days since ordering your transcript.

Rev offers both human and AI-powered transcription (the latter also powers Temi, a pick in an earlier version of this guide). It has the most modern, professional editor of the services we tested, with version control for edits and collaboration features to share transcripts (though collaborators cannot edit simultaneously). Unfortunately, it wasn’t as accurate as our current picks, with 98% accurate human transcription and only 98.6% accuracy with AI. Names were especially troublesome. Rev AI transcribed “Joel Embiid” as “Joelle and Bead,” wrote “friends” instead of “France,” and had odd issues with decimal numbers. Human transcriptions were better, but included some AI-like artifacts such as “as large as” being written as “f / f,” and meaning-changing mistakes including “caused” instead of “crossed.”

AI transcripts were worse at names but still 98.9% accurate. That’s better than all but one of the human-powered transcription services we tested, at a fraction of the time and price.

Curvedroad sign

It’s easy to access transcriptions, but editing and collaboration features are basic. Once transcription is completed, you’ll get an email with a link to a Word Document. Since you don’t need to be logged in to access the doc, you can forward a transcript to a colleague without needing to share account access.

Of other inaccuracies, “vlog” was mistaken as “blog,” “C#” was written as “CShark,” and “AI” as “AGI.” A single word—“Caucasus”—was marked as “[unintelligible]” rather than being guessed, which is what many other services would do.

Buying hours of transcription credit at a time requires slightly more upfront investment than GoTranscript, which bills directly per minute of audio transcribed, and the subscription options add to the complexity. On average, though, Vook.ai works out to around 5¢ per minute of audio transcribed, one fourth of GoTranscript AI’s 20¢ per minute cost.

Anyone who records large amounts of spoken audio and video—for interviews, podcasts, videos, meetings, dictated notes, and more—can benefit from a transcription service.

It’s geeky: You can run it in Terminal on Macs and Linux PCs, or Command Prompt on Windows. It’ll run slowly on traditional Intel and AMD processors, but it can deliver transcripts as fast as 1 minute per minute of audio on modern ARM CPUs, including Apple’s M1 and newer, or on CUDA-compatible Nvidia GPUs. Whisper’s Large v2 model was the most accurate in our tests; Small was faster, making it more usable on Intel machines, but it was less accurate, while Large v3 seemed to try too hard and overcorrected at times.

iOS 18, which rolls out to iPhones this fall, is jam-packed with AI-powered features like generative emoji, photo cleanup tools, and ChatGPT integration.

We noted where transcription services cleaned up text, removing filler words—such as “like” or “you know”—and did not count that, or when they broke up less-common compound words like “globetrotting” into “globe trotting,” or transcribing “the” as “a” against accuracy. And when the wrong word was used, we counted that against an app’s accuracy score, using that along with pricing, features, and speed to choose the best transcription service.

GoTranscript cleans up transcriptions for readability, which generally is good, though in our experience this also sometimes affected quality. Filler words including “you know,” “like,” and “kind of” were removed, and a misspoken “there’s a qualified” was turned into the more accurate “there’s a qualifier.” And you’ll have to watch out for fillers that are also modifiers, such as an instance where “kind of” was removed, which mildly altered the meaning of a sentence.

Microsoft’s AI assistant, Copilot, is now built into nearly every new Windows laptop coming out this year. Here’s what you should know about how to use Copilot.

If you need to work on transcriptions as a team: Descript (free for 1 hour of transcripts per month; from $19/month for 10 hours of transcripts) was designed as a tool for editing podcasts and videos, but it’s also a great AI-powered transcription service with a modern, collaborative editor reminiscent of collaborative notes and the to-do list app Notion.

Alice is a basic AI transcription service built around automation. Finished transcripts are emailed to you and any colleagues you add. Alternatively, the service can save transcripts to Dropbox, Google Drive, or OneDrive; share them on Slack; add them to Trello tasks; or kick off Zapier automations from completed transcripts. It whiffed on a number of names, transcribing “Cameroon” as “Kenmore,” “bobcats” as “bug cats,” and “bag” as “Baghdad,” among other mistakes that we didn’t see in other transcripts. But its built-in search was better than that of competitors, surfacing results across every transcript in your account.

For turnarounds faster than one day, GoTranscript is more expensive than our previous top pick, Rev. At $2.75 per minute versus Rev’s $1.99 per minute for the same promised turnaround time, you will pay extra for frequent timestamps or less-clear audio. But GoTranscript’s accuracy makes it the service to choose when you need the cleanest transcriptions—and actual turnaround times were faster than promised, as well.

Troad sign

In our tests, we were able to get two minutes of audio transcribed in a minute or less, with greater than 98% accuracy. The cost? Around 5¢ per minute.

None offer the accuracy of our top picks, but all offer passable transcripts that would be good enough for a quick search through meeting notes or to save time in looking up a quote.

Vook.ai supports files up to three hours long (or 2 GB), versus GoTranscript’s larger 4 GB upload limit. Longer recordings will need to be split into shorter clips for transcription, or require an Ultra plan, which costs roughly $55 per month.

It’s designed to upload audio and get it transcribed as quickly as possible. You can order transcription in around five clicks, and downloading a completed transcript takes a single click from the notification email. When ordering, you can choose AI or human transcription, then upload an audio file in nearly any popular format (including .mp3 and Apple Voice Recorder’s .m4a format, along with .webm, .ovg, and video files including .mp4 and .wmv)—or paste a link to a Dropbox or Google Drive file, or YouTube or Vimeo video.

This is not a comprehensive list of all transcription services we have tested. We have removed any services that are no longer available or do not meet our current requirements.

Vook.ai is also far easier to use than OpenAI Whisper, which requires running Python code in Terminal on Macs and Linux or Command Prompt on Windows, or using a third-party wrapper for either Mac or PC—something less technically inclined users may struggle to set up.

Amazon’s line of Echo smart speakers are the best option for bringing Alexa into your home. We’ve got picks for nearly any setup and situation.

The need for a standalone transcription service, though, has become less critical. Automatic transcripts for meetings in apps like Slack and Microsoft Teams may be enough to get general notes. Automated YouTube transcripts could suffice when starting a new channel. If transcripts are only nice-to-have for your work, a free automated solution is best.

We’ve updated this guide to correct an error. A previous version of this guide stated that Vook’s time blocks were available in two-hour increments, but they can be customized to any amount of time.

GoTranscript’s editor is basic, but the service includes a number of features that can speed up your work. You can get a free AI transcript preview of the first two minutes of your audio, which allows you to decide if the quality is sufficient before paying—and if it’s not, you can upgrade to a human transcription. You’ll get your transcript emailed as a Word document, so you can use it without having to re-open the site. And it’s HIPAA compliant, so it’s ready for health care use.

GoTranscript’s site mentions that it “uses AI as a tool to enhance the efficiency of human transcriptionists,” which may be part of the reason its human transcripts are so accurate and delivered so fast. To improve accuracy, you can email the GoTranscript team to add industry-specific flags to your account for medical, technical, and legal terms, or custom words to add to your account’s dictionary. It would be nice to have an in-app dictionary feature to add custom words on your own, which is something offered by many competing services, including Scribie and Rev.

It uses OpenAI’s Whisper to generate fast, accurate transcriptions. What makes Vook.ai one of the best affordable transcription services is that, at its core, it’s an easy-to-use take on OpenAI Whisper. (According to the development team, it’s powered by “added layers upon OpenAI’s Whisper algorithm, that we run on serverless GPUs.”) The results were slightly more accurate than default Whisper models in our tests.

GoTranscript provides transcripts with over 99% accuracy from human transcribers in less than a day, at more affordable rates than many human-powered alternatives. Its custom AI transcription is also the most accurate automated transcription service we tested, as well.

The results are only slightly less accurate than what you’d get from our top pick, but enough that you’ll need to more closely edit transcriptions before publishing them. It also offers no collaboration features, requiring workarounds to handle transcripts as a team.

Interestingly, the opening pangram in the final audio clip was merged together as a single word, just as in GoTranscript’s AI copy, leading us to wonder if the same core models underlie GoTranscripts’ AI transcription.

The service gives you a preview transcription of the first couple of minutes of your recording for free, to check its quality before paying for a full transcription. And it can import nearly any audio or video format, as well as export transcripts in a wide range of formats, to slot easily into your existing production workflows.

Getting transcripts is easy and quick. Vook.ai transcribes audio three times as fast as GoTranscript’s AI service. You simply upload recordings from your computer, choose to optionally identify speakers (free) or translate the transcript with DeepL (for double the cost).

Otter is built around transcribing meetings, with Zoom, Microsoft Teams, and Google Meet integrations. Its best use case is to surface data from calls, with automated AI summaries of transcripts, a chat bot that you can ask questions about a call, and graphs tracking who talked the most. It’s less useful for working with transcripts, though, with a limited editor that only lets you copy a single paragraph at a time. It offers few export options and is also somewhat less accurate than the competition.

Trint is designed around transcribing large volumes of audio every month. It offers unlimited transcription starting at $100 per month, in a collaborative editor complete with full-text search, sharing, and comments. That is helpful for transcribing large volumes of audio for subtitles, with quick AI transcription that you can clean up manually in the app. It includes a full-featured mobile app so that you can get text transcribed on the go, and it supports exports in a wide variety of subtitles. Unfortunately, it was just 95% accurate in our tests, far off our picks. It also made errors that changed the meaning, like transcribing “steady” as “study,” “nebulously” as “nebulous Lee,” “That’s always gonna” as “I was always going to,” “right now” as “if I know,” and more—not close enough to publish without a second check.

Its transcripts are 98.7% accurate on average. In our tests, Vook.ai was second only to GoTranscript in AI transcription accuracy. The results were good enough to use with only light edits.

Automated transcription services can transcribe in nearly real time. Human transcription services should take no more than one business day—and ideally less.

Bridge aheadsign

iOS 18, which rolls out to iPhones this fall, is jam-packed with AI-powered features like generative emoji, photo cleanup tools, and ChatGPT integration.

Its human transcriptions were over 99% accurate. The words the human transcribers missed were more likely to be filler words—for example, writing “while” instead of “and while”—rather than entirely incorrect words. Names that were frequently misspelled by other transcription services (including “Sagen Maddalena,” “Will Shaner,” and “Joel Embiid”) were transcribed accurately. Transcripts were close enough to perfect that we had to look closely for differences.

Collaboration is similarly basic: You can copy a sharing link to the transcription editor, where colleagues can edit the text but not actually save the edits. Or, you can add transcripts to a shared workspace for team access with the same basic editor features. You can filter through transcripts by date and order but cannot search through all recordings together.

Matthew Guay is a writer focused on software and productivity. Previously he was a writer for the automation platform Zapier and a founding editor of the software community Capiche. With more than 1,500 logins in his password manager, he has lost track of how much software he has tested.

Transcription Panda is a human-powered transcription service with the most accurate transcription we tested—but it costs $1 more per minute than GoTranscript with one-day turnaround and takes nearly six times as long to transcribe the same audio. It’s also very basic: You just upload audio and get a Word document emailed back to you. There’s no sharing, editing, or playback features. The few errors the service did make were more likely to change the meaning of a sentence than those from GoTranscript’s human transcripts, too; “instance” was transcribed as “insurance,” for example.

We compared each service’s interface and listed core features, including supported audio formats and collaboration options. And we timed how long each service took to provide transcripts and respond to support queries, and noted the total cost per minute.

WindingRoad sign

It missed one name (“Sagen” was transcribed as “Sajan”) and used British spelling for “catalogue,” with no option to choose between US and British spelling. It also had a few issues with our final test clip: The pangram was written without spaces, “AI” was written as “AVI,” and “streaming” had become “steaming.” Perhaps most surprisingly, one clip had a phantom “Thank you” added to the end.

With the transcripts in hand, we then compared each transcript against our original, hand-transcribed copy to compare for readability and accuracy. Our primary criteria is accuracy in transcribing spoken words; that’s the sole feature worth paying for.

But when accurate transcripts are crucial enough to your work that you would otherwise spend hours painstakingly transcribing or editing automated transcripts, paying for human transcription or higher-quality automated transcripts is the best choice. Paid transcription tools also include features to share transcripts, edit them manually, and export the text in caption and subtitle formats. All of these are worth paying for when transcripts are a core part of your work.

Descript is worth considering for its interface alone if you’re working with a team. And if you’re building transcripts for a podcast, it could be an all-in-one tool to record, edit, and transcribe podcasts as a team.

At a quarter of the cost of our top pick, this service offers marginally less accurate AI transcripts that arrive in less than half the time. But it has fewer features and can struggle with crosstalk.

GoTranscript is the best transcription service for highly accurate human transcription, and it also delivers better-than-average AI-powered transcripts. It’s worth paying for when transcription accuracy is crucial.

That has improved their accuracy but can lead to surprising mistakes. For example, in our testing “as large as” was transcribed by multiple AI services as “F/f.” Even more strangely, the decimal “1.82%” was transcribed as “1.8199999999999998%” by one service. Yet overall, today’s AI transcriptions are 96% accurate on average. Close, but still noticeably incorrect at times.

It keeps transcripts in-house. User data is stored on EU-based servers and RSA encrypted, and transcriptions run on an OpenAI Whisper model run on Vook.ai’s infrastructure. “Audio files or transcripts are not shared to any third party,” the Vook.ai team told us, unless users request DeepL translation or GPT-powered transcript summaries. Automated transcriptions don’t require a human to listen to audio or read the transcript, keeping transcription private by design. It’s not HIPAA certified, though, which makes GoTranscript a better choice for medical and legal use-cases if compliance is required.

GoTranscript is safe to use with sensitive personal information. The service uses in-house–developed transcription software, promises limited access to personally identifiable information to employees and transcribers who need access to data to perform specific tasks, is HIPAA-compliant, and encrypts data while stored and in transit. Audio is stored for seven days after transcription is complete, while text transcripts are stored in your account until you delete them. If needed, you can also request the GoTranscript team to sign an NDA for an additional layer of privacy around your audio and transcripts.

And the transcriptions are good. Surprisingly so. In 2018, Wirecutter found the best AI transcription tools to be 73% accurate. Today, even the least accurate AI transcription is 94% accurate—and, surprisingly, the best AI transcription services, including OpenAI’s free Whisper transcription engine, are somewhat more accurate than the least-precise human-powered transcriptions.

Vook.ai is a sparse transcription service, built using custom transcription models based on OpenAI’s Whisper. But it delivers highly accurate transcriptions in minutes.

We submitted these clips to each service, testing both AI and human transcription options where possible. With human transcription services, we added technical terms from the fourth script to the services’ custom dictionaries and paid for one-day turnaround time.

For AI transcripts, GoTranscript doesn’t prompt you to add any additional services, as AI transcriptions list speakers by default but don’t offer timestamps. As a nice bonus, you can click a “Check AI Accuracy” button when ordering an AI transcript to get a quick preview of the first two minutes of your audio’s text. If the results aren’t accurate enough, you can upgrade to human transcription.

Alternatively, you can use it in third-party wrappers, including MacWhisper for Mac, Aiko for iOS, and Speech Transcribe for Windows PCs with CUDA-compatible Nvidia GPUs. Or, you could use hosted versions of Whisper via automation tools like Zapier or Make, or build it into in-house apps with OpenAI’s Whisper API.

Although its mobile app is only usable to record conversations to then be submitted to GoTranscript, and its web editor includes fewer features than some competitors, its accuracy and speed are the best we’ve found.

You can also choose your timeframe: Six- to 12-hour transcription turnaround starts at $2.75 per minute, with one-day ($1.43 per minute), three-day ($1.21 per minute), and five-day (99¢ per minute) options also available. In our testing, we used one-day turnaround with default options, plus timestamps for two of the clips. This worked well, even with accented, noisy audio.

At a quarter of the cost of our top pick, this service offers marginally less accurate AI transcripts that arrive in less than half the time. But it has fewer features and can struggle with crosstalk.

State routesign

GoTranscript uses AI to enhance its human transcription, and that shows in the year-over-year improvement in its transcription quality. The web app is fairly basic, with simple sharing functions, plus a text editor with audio playback that highlights words as they’re spoken.

Similarly, Apple Intelligence—available starting September 16 in iOS 18 and macOS Sequoia, on iPhone 15 Pro and iPhone 16 and newer, and iPad or Macs with M1 CPUs or newer—generates real-time transcripts from audio recordings in Notes with AI summaries. We wouldn’t be surprised to see direct audio file transcription for both platforms in a future update.

In the same way, OpenAI’s Whisper transcription model is available in code today, or via an API call on OpenAI’s servers. While ChatGPT today only includes dictation used to ask a question, it would seem to be a natural progression to offer audio file transcription directly in ChatGPT’s interface—or to see similar offerings from other LLM chats such as Claude in the near future.

Apple Voice Control and Nuance Dragon Home 15 are the best dictation tools we’ve tested, though both struggle with some accents and speech impediments.

Reduct is an online video editing app that transcribes audio, too—a video take on what Descript does for podcast editing. Transcripts show up in real time as the AI completes them, with a quick first-pass (taking one minute per minute of audio) followed minutes later by a more accurate final transcription (around three minutes per minute of audio in our tests, which is slow for AI transcription). It missed an entire phrase on the end of one clip and added extraneous phrases to the end of two other clips. Reduct’s nicer editor and collaboration features could make it worth considering if you’re primarily transcribing videos then publishing the copy as captions, as long as you edit transcriptions before publication.

Unlike our top pick, there are no sharing or collaboration features, though Vook.ai does have an email export option to send a transcript directly to a colleague. Similar to GoTranscript, though, Vook.ai doesn’t offer universal search across all of your transcripts (something offered by Alice, Rev, and Descript). The only way to organize transcripts is by their read or unread status, or with tags.

If you want free, locally processed AI transcripts: OpenAI Whisper—the engine behind Vook.ai and likely other commercial transcription tools—is an open-source, MIT-licensed, Python-powered transcription model that you can run for free on any computer. Its Large model was 98.7% accurate in our tests, while its faster Small model was 97.7% accurate.

I’ve been testing and writing about software professionally since 2010, and I have used both human and AI-powered transcription services for interviews and meetings throughout my career. I extensively use dictation for mobile typing, across multiple languages, and regularly use transcripts to build note summaries after meetings.

A couple of names were misspelled—slightly more than with GoTranscript; “Roe” came back as “Rowe,” “Bhupinder” as “Bupinda.” It didn’t handle crosstalk as well as our top pick and missed a few smaller words including “did,” “they,” and “yeah.” Percent was written out as a word, instead of a percent symbol.

Scribie is the most affordable human-powered transcription service we tested, at just 80¢ per minute. It first uses AI to transcribe your audio, then human editors clean up the copy. Scribie was as accurate as Happy Scribe, though with a few more spelling differences than other transcription apps. (These included things like using “okay” instead of “ok” or “all together” instead of “altogether.”) It also produced near misses like “study” instead of “steady.” It offers fewer export options, with only Word, PDF, and text formats supported, with no subtitle or caption formats included. It was very good at handling crosstalk and repetitive or filler words, though.

At a quarter of the cost of our top pick, this service offers marginally less accurate AI transcripts that arrive in less than half the time. But it has fewer features and can struggle with crosstalk.

If you want free, good enough transcripts: Check your favorite work applications to see if they include transcription features. Microsoft Word includes automated transcription with Office 365, Slack adds a transcript to any audio shared in chat, and most video call apps including Google Meet and Zoom offer automated transcripts.

On the other hand, AI transcription tools today rely on neural network transformer models that recognize words in spoken audio then infer the correct words and punctuation. Similar to the way predictive text on your keyboard recommends the next word, or ChatGPT gives answers, today’s transcription tools are, at a high level, guessing what you’re going to say next.

Today’s smartphones already offer basic transcription features, typically for voicemail or as accessibility features. Google Recorder provides high-quality real-time dictation, and Google’s Gemini can create a text summary of recorded audio on Pixel 8 and 9 series phones.

Beyond that, the transcripts were clean and readable, and the AI handled crosstalk, repeated words, and other verbal issues well. The results are close enough to perfect that, for clear audio without uncommon words, you might be able to use them without any edits.

After spending 35 hours testing 15 transcription tools, including human and AI-powered options, we found that GoTranscript is the best service for those who need the most accurate transcripts. If you’re willing to work with slightly less accurate but far faster AI transcription, Vook.ai is the best affordable transcription service.

For this guide, we considered how the field has changed since we wrote Wirecutter’s original transcription services guides in 2018. Then we read through personal recommendations from New York Times reporters, along with roundups from The Verge and PC Mag. We also took into account recommendations from NYU Libraries, the University of Michigan Library, and Harvard University, among others, as well as forums and social media.

For human transcription, you can then choose between clean verbatim (without filler words such as “um” and “you know”) or full verbatim for an additional 25¢ per minute. Additional paid add-ons include timestamps, support for three or more speakers, and support for low-quality audio.

Vook.ai is a highly accurate, AI-powered transcription service with one of the lowest prices we could find. On average, you’ll pay 5¢ per minute of audio transcribed—and will receive completed transcripts in less time than the duration of the audio recording, with speaker labels and timestamps.

Apple Voice Control and Nuance Dragon Home 15 are the best dictation tools we’ve tested, though both struggle with some accents and speech impediments.

Your first 30 minutes of transcription are free, then you can purchase credit in two-, five-, or 10-hour increments, or customized blocks of time. Another option is to purchase a subscription starting at around $11 per month for five hours of transcription, with transcription billed per minute and charged in euros. Leftover credit can be used later.

Transcripts show up in your account moments after they’re ordered. The service includes a basic editor, with rich text editing features including bold and italic font options. You can toggle timestamps and speaker names, copy text, or export transcripts in Word, PDF, Markdown, or .st subtitle format.