Education How does automated transcription work?

Automated transcription has come a long way in the last few years. Whether you’re a podcaster, journalist, video editor, lawyer, student, or researcher, the need to convert audio or video to text is a part of your life. Transcribing audio and video, lightly put, is painful.

But now there is technology like Sonix that automates the entire process of transcription. Hence the term automated transcription.

 

 

What is transcription?

Transcription is the process of converting recorded speech into text. Transcription involves listening to a recording of something and typing the contents into a document. In many cases, this is an interview of some sort. It can take an inordinate amount of time to transcribe something especially if you are doing it manually.

 

What are the different kinds of transcription?

Manual transcription

The most traditional form of transcription is manual transcription. Manual transcription involves listening to audio or video files and then typing the words into a document. Many people choose this option because there is no associated cost. The cost equates to much an individual values their time.

Human transcription services

There are many human transcription services on the market but they can be slow and expensive. They are, however, more accurate than automated services. Human transcribers use technology to assist in the conversion of speech to text like a shorthand system. There are also a very small number of people like court reporters that can type in near real time. The accuracy of these transcripts is negatively affected because there is little to no time to correct mistakes as they occur.

In summary, human transcription has been around for decades, it isn’t the most efficient or effective way to convert audio or video to text.

Automated Transcription

Automated transcription, as the name suggests, is amazingly fast. While human transcription can take anywhere from 24 hours to 4 days, automated transcription can be completed in minutes. A 30-minute audio or video file can be transcribed in less than 5 minutes.
Another added benefit of automated transcription is security. No human ever sees the audio or video file, nor the transcript. It’s done entirely by machines. If the security of your files and transcripts is important to you, then automated transcription in many ways is better than human transcription.

Sonix uses the latest artificial intelligence and natural language processing techniques to derive the most accurate automated transcripts. Sonix has been independently reviewed as the most accurate automated service.

Once a file has been transcribed, you’ll receive an email notifying you that your transcript is ready. All the transcripts are centrally hosted in your Sonix account online for easy access. Just click on the link and you will see the time-coded transcript displayed in your browser. Because your transcript is online, Sonix stitches the audio to the text which makes it easy to edit your files. You can also easily search for keywords, share the file with another user, highlight and strike text, embed captions in video, and export in many different formats.

To be clear, automated transcription is not perfect. The technology continues to get with every day that passes, but there will undoubtedly be errors. And if your file is recorded in a loud environment, there are people talking over each other, or speakers aren’t articulating clearly, the resulting transcription will be negatively affected. On the other hand, with really clear, crisp audio, the accuracy of the transcript can be upwards of 95-98%.

Lastly, automated transcription is relatively inexpensive. While traditional human services can cost anywhere from $60 to $100 per hour of audio or video, automated transcription with Sonix is just $6 per hour with a subscription. The effective cost for those that transcribe regularly is substantially less.

 

What makes automated transcription possible?

Automated transcription is possible because of artificial intelligence and natural language processing. With each file that is uploaded to Sonix, each and every sound within that file analyzed and interpreted using artificial intelligence and natural language processing.

The next step is to match those sounds to words in our extensive and growing dictionary. Sonix works in different languages and varying English accents and continuously improves as more voice data is ingested into our systems.

There are four basic steps with automated transcription:

STEP 1

A user uploads an audio or video file and selects the appropriate language spoken (Sonix works in many languages and several different English accents).

STEP 2

Sonix then runs the file through its automated technology which combines artificial intelligence and natural language processing to derive accurate transcripts. Depending on the size of the file, this can take anywhere from 2 to 8 minutes.

STEP 3

When the file is complete a user will receive an email with a link that takes them to the online transcript.

STEP 4

With a browser-based transcript, users can easily edit problem areas.
 

The future of automated transcription

We are several years away from error-free automated transcription, but the technology continues to improve day after day. As more and more people turn to automated transcription, more voice data is collected and analyzed. The result is improved speech-to-text algorithms and more accurate transcripts.

In the meantime, there are many ways to get the most out of automated transcription in its current state. Most of that requires users to capture high quality audio and video. Reducing background noise, multiple speakers talking over each other, and swallowed words can greatly increase the accuracy of automated transcription.

 

New to Sonix? Click here for 30 free transcription minutes!

Education What are the benefits of using multitrack in Sonix?

Why use multitrack?

Hi Sonix users. Today we're going to talk about how to use multitrack in Sonix. I've got Bill here to explain some of the benefits. Hi Bill.

Hey Adam great to be here.

Bill can you tell us the benefits of uploading multitrack files versus singletrack in Sonix.

Sure. Many of us capture each speaker separately by using different microphones to isolate each person's voice. The result is a multi-track recording which has multiple benefits when using Sonix.

Ok what are those benefits?

Automatic speaker labels

The first is that you can name the speakers before you upload. This gives you near perfect automatic speaker labeling which saves a ton of time going through and naming speakers after the transcript has been returned to you.

Nice. That's a good one.

More accurate transcripts

Yes. And secondly you'll get more accurate transcripts because the voices on each track are isolated.

That's great too. Why wouldn't everyone use multitrack

Always capture speakers on multiple tracks if you can

While we always suggest capturing multiple speakers on different tracks. There really isn't any good reason to have multiple speakers on one track.

Great thanks for this bill. You are awesome.

I know that.

Automatically convert audio to text with Sonix



–>

New to Sonix? Click here for 30 free transcription minutes!

Sonix Tutorials Introduction to Sonix

Just click on the player below to view the Introduction to Sonix.

You can also click on the burger icon to navigate quickly to key parts of the tutorial.

Sonix Tutorial – Introduction to Sonix

How to upload a file in Sonix

How to upload a file a Sonix

To upload a file to Sonix you simply click on the upload button. Sonix accepts multiple file formats listed here for both audio and video.

Sonix also accepts the following languages.

You can upload a file by dragging and dropping the file into this quadrant, selecting it from your Dropbox account, or your Google Drive account, or simply selecting it from your desktop which is what I'm doing here.

The green status bar shows the status of your upload. You can actually upload multiple files. I'm just going to add the same one again. You can see that this is uploaded as well but I've decided I don't want this file. So I can easily remove that here.

You can rename the file and then you're ready to transcribe.

The last thing is to make sure that language that you select is the language spoken in the audio or video file. In our case it's English American so we're good.

And then you just hit start transcribing.

From there you're taken back to your home screen. And you can see that this file appears. And the status is also here. Right now the file is preparing. It takes roughly five or six minutes for an hour of audio or video to transcribe. And once it's done you'll be notified from Sonix.

How to edit a file in Sonix

How to edit a file in Sonix.

Just click on the file and you'll be taken to the transcript.

It looks a lot like a word processor but the big difference with Sonix transcripts is that the audio is switched to the text. If you click anywhere in the transcript you can hear the voice behind the words that are spoken.

My name is Pete Combs and I'm a reporter for K O M O KOMO Radio in Seattle. I've been doing this for about 40 years and my life has.

This makes it super easy to edit your transcript and correct any errors.

Now there are tons of features around editing for your transcript. But in this tutorial I'm only going to go through 5 key thing. Number 1: using Heatmap to locate problem areas, 2: Sonix shortcut keys, number 3: Find and Replace, number 4: creating a new speaker, and 5: highlighting and strikethrough.

Using Heatmap to locate problem areas

Using Heatmap to locate problem areas

If you click on the tiny thermometer Sonix will highlight the areas that we aren't confident or correct. The lighter the text the less confident we are and the darker the more confident we are. This allows you to navigate around the transcript and quickly fix up areas that look more problematic.

Shortcut keys in Sonix

Shortcut keys

If you click the shortcuts menu item in the top now it shows you all the shortcut keys that can help you speed up editing.

An important shortcut key is TAB. TAB allows you to stop and start the audio while editing.

Find and replace

Find and replace does exactly what you'd expect it to do. You can find it under the edit dropdown or you can simply click this icon.

Enter the word you want replace. The word that you want it replaced with. Click replace all and you'll see a notification that shows how many times that word was changed.

Creating a new speaker

Creating a new speaker

If you have files with multiple speakers you may run into issues where we don't separate speakers perfectly. That's not a problem. You can easily create a new speaker by clicking anywhere in the text.

Say this was a new speaker here. Just click enter and create a new speaker by entering here. It's as simple as that.

Highlighting in Sonix

Highlighting and strikethrough

Highlighting and strikethrough are two features that are mostly used to prepare your file or sections of your file for export. To highlight word, sentence, or paragraph, simply click on a word and hold and drag until you've captured the section you want.

Then simply select the paintbrush. You'll see the highlight turns yellow and Sonix also identifies the time stamps for the start and the end. You can also see the highlight in the progress bar so it's easy to navigate.

Strikethrough in Sonix

Strikethrough works much the same way as the highlight. Simply select the area for the paragraph you want to strike and click the strikethrough icon. The text that's been struck through is highlighted in red and it also appears the progress bar.

Organizing your files in Sonix

Organizing your files in Sonix

Your Sonix account operates in a similar way to Dropbox or Google Drive. We store all your audio video and transcripts. It's all part of this subscription fee.

You can create update delete and organize your files in any way you want. The first folder is called "Shared with me". This folder keeps all the files that have been shared with you by someone else.

If you want to create a new folder simply select new folder. You can rename move or delete your folder. And you can drag and drop files inside of other folders.

How to share a file in Sonix

How to share a file in Sonix

You can easily share a file with another person so they can view the transcript. Just click share.

Here you have two options. You can copy this link and send it to anyone. But note that the recipient can only view the transcript.

You can also enter someone's email address. And select whether the recipient can view only or edit as well. There is also space here to write a personal note.

Just click invite and they'll be invited to the transcript.

How to export a file from Sonix

How to export a file from Sonix

Exporting a file is simple. Just navigate to the file you want to export and click export.

You can export your transcript audio or video in a variety of different ways. This makes it easy to get into your workflow quickly.

First you can export the transcript in text format into Microsoft Word or as a text file. And then you have several times stamp options. Include timestamps on every paragraph, include timestamps every 30 seconds, include milliseconds with the time code.

You can also export with or without a speaker names.

And lastly you can export just the highlighted section.

You can also export the transcript into audio and video editing software like Adobe Audition, Adobe Premiere, and Final Cut Pro.

If you want to include captions alongside your video you can do that by exporting an SRT file.

Lastly you can export the audio from your file. And you have three options here. You can export the whole file, just the highlighted section, or the audio with the strikethroughs removed.

Automatically convert audio to text with Sonix


New to Sonix? Click here for 30 free transcription minutes!

Review Should I go to Podcast Movement? -> A quick review of Podcast Movement

Attendees are bullish on the future of podcasting

We just finished up Podcast Movement in Philadelphia. An event where podcasters, podcast producers, and radio folks convene to talk and learn about everything related to podcasts. We loved meeting face to face with many of our customers and meeting a bunch of new people too. There are so many great stories being told and so many interesting projects among this community. For many attendees podcasting is their livelihood. For others, it was another ‘must-have’ marketing medium for their brand or a place to learn how to start a podcast. One thing was incredibly consistent: everyone was bullish on the future of podcasting.

            

 

What’s the structure of Podcast Movement?

Much of what you’d expect happened at Podcast Movement including keynotes by industry influencers like Pay Flynn, Jarl Mohn, and Terri Gross. Google made an appearance to talk about their new foray into podcasting. There were break-out sessions every day on topics ranging from marketing to production to monetization to ‘how to create a podcast’. But what we found made the conference great was the hallway conversations. So great to meet loads of geniunely great people with passion and interesting stories.

 

    

 

Things outside the event
Philadelphia treated us well. The Reading Terminal is an awesome place to grab lunch; it’s filled with every kind of food you could imagine strewn with stalls from end to end of the iconic building. We lucked out with our dinner choices too. Talula’s Garden was our favorite. If you’re going to Philly make a reservation. It won’t disappoint.

 

Should I attend Podcast Movement in 2019?
Every individual and company is different. We base our decisions on a mix of ROI and exposure. We found it a good use of our time and plan on going in 2019. See you there!

 

Education How to fix mic bleed with multi-track recordings

If you are using Sonix’s multi-track feature you may experience an issue called “mic bleed”. Mic bleed happens when you can hear two or more voices on one track.

Isolating audio tracks is challenging especially if you have multiple microphones in the same room. Sonix has built technology to help with mic bleed. We can spot areas of duplicated text and remove them from the combined transcript. Having said that, if the mic bleed is loud, Sonix will struggle to isolate voices.

Mic bleed is less problematic when you record speakers in different rooms or locations. It used to be a logistical challenge to capture each participant on their own audio track, but with all the great communication apps, it’s no longer a problem. If you are using Skype, Zoom, or some other tool to capture the audio, each speaker is using a different microphone input. This helps reduce mic bleed.

Can I fix mic bleed?

There are ways to strip unwanted audio from your tracks using other audio tools like Audacity, Final Cut, and Pro Tools. The feature is typically called “Strip Silence” which will detect areas in an audio track that are quieter than others.

Here are the links to instructions on how to strip out mic bleed.

  1. Audacity
  2. Apple Logic X
  3. Pro Tools

New to Sonix? Click here for 30 free transcription minutes!

 

Announcements Multitrack transcription and automatic speaker labels

If your audio or video recording has speakers on separate tracks, you are going to love our new multi-track feature! You can upload each track separately and then Sonix will combine those tracks into one transcript.

Why use multitrack?

Many podcasters and radio folks capture each speaker separately by using different microphones to isolate each person’s voice. The result is a multi-track recording which has multiple benefits when using Sonix:

    1. You can name speakers before you upload to get automatic speaker labeling with near-perfect precision
    1. You will get more accurate transcripts because voices on each track are isolated
  1. You can export from Sonix and retain separate tracks

You can set up multi-track upload in your account under preferences: http://my.sonix.ai/account/profile/multitrack

New to Sonix? Click here for 30 free transcription minutes!

Announcements Laurel vs Yanny: AI speech recognition leaders Google, Amazon, Sonix, and IBM can’t agree either

Sonix, an emerging leader in speech recognition, compared its speech recognition engine to the biggest players in this space including Google, Amazon, and IBM.

The results are mixed with Google & Sonix recognizing “Laurel”. Amazon recognizing “Year Old” which is closer to Yanny. And IBM recognizing both “Yeah role” and “Laurel” which is like a mix of Yanny and Laurel.

To conduct the test, Sonix used the Yanny vs. Laurel version that Cloe Feldman first posted on Twitter.

Here are the results:

Google’s Speech API

Amazon Transcribe

Sonix

IBM Watson Speech to Text