Which Transcription Service to Use for Each Job [updated]
With audio transcription, you get what you pay for. For a dollar per minute, Audio Rev will provide you with a well-formatted transcript. For 25 cents a minute (a little less for high volumes) Trint provides you with a transcript riddled with errors, but helps you fix it with a sophisticated editing tool. And for 10 cents a minute, the all-new CheapTranscription.io gives you a simple text document with even more errors. You can use all three depending on your needs. Here’s how to choose.
For public transcripts: Rev
Let’s say you are producing a podcast with multiple voices and you want to publish the transcripts of each episode. Or tell me you want subtitles for your videos. Or say you interviewed someone and want to print the entire interview. If you can afford it, Rev is the best solution.
The price is high
Rev charges $ 1 per minute for transcripts and captions for English videos. Video captions include time stamps.
It adds up quickly if you transcribe a weekly hourly podcast out of your pocket, but it’s cheap enough for a corporate budget. I’ve also used Rev to create high quality titles for my personal creative video projects. There is no minimum fee, so I even ran 3 minute YouTube videos via Rev.
These are flat rates, with no additional fee for more speakers. You can pay extra, for example, to include each verbal tic or to include time stamps in the transcript. Rev also offers translations of written documents at 10 cents per word, audio and video recordings at $ 3-7 per minute.
Transcript quality: high
Upload your audio or video to Rev and they’ll be back within 24 hours, with meticulous transcription, every line labeled by the speaker, and “ums” and other word tics removed. Any illegible words will be marked. Usually they are minimal, from two to five per hour of telephone conversation.
Rev decoders can accurately record both sides of a telephone conversation or face-to-face interview with medium background noise. They can work with a fairly wide range of accents, although if you find it difficult to hear something in your audio recording, they will probably have problems too. They are not magical.
Here is an excerpt from a telephone interview I did with Welcome to Night Vale actor Cecil Baldwin , transcribed by the Reverend:
Video subtitle transcription is accurate and timed, which is very important – sloppy subtitles stand out immediately in online videos. At the Rev rate, it’s always worth paying to transcribe any short to medium video, especially if you’re uploading to platforms like Facebook or Instagram where a lot of people watch the video without sound. You will save yourself a lot of rough work.
Publisher: Mediocre Great
Transcription service is never perfect. So in most cases, you need to do at least a very light cursory scan to check the names and any insertions [inaudible] or [inaudible] , or some overly accurate transcriptions of the interviewer inserting “right” and “Oh!”
Rev provides its transcripts in an editor interface that includes an original audio player. In the editor, you can directly edit the text, change the names of speakers, annotate certain passages, and share the transcription with other users. You can export text in docx, PDF or txt format.
Update: March 27, 2019 9:30 AM ET: Rev recently made a major update to their transcript editor. Like Trint, Rev now verbatim matches the original audio to the transcription, so you can easily correct mistakes. This change is not retroactive, so any transcripts you ordered prior to the update will still use the old editor described below. The new editor makes Rev a much better option than before.
Previously:
Unfortunately, the Rev audio player does not sync with the cursor in the text and does not highlight the corresponding text when playing a sound (at least in our tests in Chrome). So, if you need to find a specific line that you need to fix, you have to search the record. This is an unfortunate oversight on the part of a high-class service.
When You Don’t Need Perfection: Trint
If you need to dig into a recording but don’t really need to publish it (or have time to edit it), you don’t need to pay the high rate for human transcriptions. Machine learning is very helpful in speech recognition, and while automatic transcription services are not as accurate, they are good enough for many purposes and are much cheaper. Our favorite and the one we use at Lifehacker is Trint .
Trint was founded by war correspondent Jeff Kofman ( whom I interviewed for Lifehacker ) and is calibrated to be useful for a variety of internal transcription needs. This is a great solution for journalists who need to search for a specific quote in a recording, or who want to publish a partial transcript, or want to create a full transcript in less time.
Price: Low
Trint charges 25 cents per minute of audio or video, which is a quarter of Rev’s rate. And because it is automated, the processing time for the transcription is usually less than an hour. Like Rev, Trint charges a flat rate regardless of multiple voices or accents, and includes additional timestamps.
If you transcribe more than 3 hours of audio every month, you can subscribe for a slightly lower price of 22 cents per minute. With 10+ hours per month, you only pay 20 cents per minute.
The subscription includes use of the excellent Trint transcript editor, which is a killer product.
Transcript Quality: Medium
At these rates, transcribing the weekly podcast is relatively affordable, but it would take some extra work. Trint’s automated transcripts are nowhere near as accurate as anything human-made. They easily go astray on phone calls. But if you have a live recording of every side of the conversation – even with cheap microphones – then Trint will make the recording mostly correct.
Here is Trint’s transcript of Cecil Baldwin:
One shame that Trint does catch every verbal tick. And when you read the transcript, you might notice that a lot of smart, eloquent people are filling their speech with little um and the like , and you know the s. I recently interviewed Stephen Wolfram, creator of software like Mathematica and WolframAlpha , one of the smartest people I have ever talked to, and I am currently editing the transcript at Trint. I had to delete annoying insertion words in every sentence he uttered. Barely noticeable in speech, very noticeable in transcripts. (Actually, a verbatim transcript is a great way to make someone seem stupid, so a little editing is an ethical issue.)
Editor: Excellent
Trint Editor is the real advantage of the service. Place the cursor anywhere in the transcript and you can start playing audio from that exact spot. It allows for fast and accurate editing and helps to interpret the many confusing incorrect transcriptions that Trint does.
The editor allows you to mark paragraphs as you complete them, add bookmarks and comments, and train Rev in specific words, names and phrases. There is also a find and replace function. There are detailed playback controls with keyboard shortcuts, and once you learn them, you have a smooth workflow with audio easily attached to the written word. You can export text as docx, xml, edl for Premiere, HTML player, embeddable Trint player, or srt or vtt subtitles.
Even if you want a clean, post-worthy transcript, you’ll spend far less time editing a Trint transcript than manually transcribing audio from scratch. You will need to measure your productivity, but you will likely pay much less for the time saved than your own hourly rate. And if you can use the organization’s budget, it’s worth it. This is why Lifehacker’s publisher, Gizmodo Media Group, bought a full subscription to Trint for their reporters.
When you just want the gist: CheapTranscriptions.io
Maybe you need a “good enough” transcript to get the message across. You don’t give a damn about editing software and don’t mind if every sentence contains at least one wrong word. Perhaps you just need a link so you can manually view the audio interview for a specific quote. Perhaps you just want your podcast episode to appear in relevant Google search results. If that’s the case, you can get by without using CheapTranscription.io , a new quick and dirty auto transcription service from tech journalist and podcaster John Biggs and developer Tom Printy.
Price: Very low
CheapTranscription is really cheap: charged 10 cents per minute with a minimum order of 50 cents. Upload your audio (in our test, the site did not accept video), and in a few minutes you will receive a text transcript.
There is no export function right now, so you’ll have to copy and paste your transcript into a text editor. And there are no time stamps. The speakers are divided but given common names.
Transcript quality: low
Oddly enough, CheapTranscription seems to be making more mistakes than Trint, as you would expect from a new service. Like Trint, it includes all verbal tics, which makes reading the transcription very tedious. But you can still grasp the essence of the passage without any editing, and you can usually tell which words are errors.
Editor: No
Your transcript is just text on a static web page. Edit in Word or Google Docs or wherever you usually write.
And that’s the real problem: Trint helps you fix it. If you want to edit CheapTranscription’s transcription, you will still want to use a specialized transcription editor like oTranscribe or Otter .
You will still get better results than the free option: stream the audio over YouTube and get a transcript. At this point, you are approaching nonsense.
This service is two days old at the time of posting, so expect to see more features as Biggs and Printy evolve. They have already announced plans to add human decryptors. Also expect an improvement in transcription quality as you learn your algorithms. Of course, this does not mean that they will ever catch up with Trint or other algorithmic services.
Buy only what you need
Rev and CheapTranscription have virtually no minimum ordering, while Trint offers all of its best features for $ 40 / month. This way, you can place orders for all three services as needed.
In most cases, Trint is enough for me. (If I paid out of pocket, I would use it less often, but still prefer CheapTranscription.) But when I run out of time to process a lot of transcripts, I ask my editor to shell out money for the Rev. Human transcriptions cost four times as much, but they are incredibly accurate. Last year, my editor and I used Roar to prepare transcripts of four telephone interviews, which we then edited into one long article on the Welcome to Night Vale touring audio drama . Getting accurate transcripts saved us hours of work on a high-profile piece.