
On February 6th, Accessibility NYC hosted a meetup at the Andrew Heiskell Braille and Talking Book Library in New York. At the meetup, Bryan Gould and Geoff Freed from the National Center for Accessible Media gave a presentation about CADET, their new (and free!) tool for writing captions and descriptions.
Here are some highlights from the presentation:
- What NCAM does
- Captioning a video with CADET
- Positioning and styling your captions
- Writing audio descriptions in CADET
- Writing descriptions vs. writing captions
Transcript: CADET, NCAM's New Caption and Description Editing Tool
THOMAS LOGAN: Welcome, everyone, to Accessibility New York City. This is our first meetup of 2018. Great to have so many people in attendance tonight. Our goal and one of the things we strive to do is to have, you know, a hightech, livestream meetup where we have live captioning in realtime, as we present. My name is Thomas Logan, I'm one of the organizers of the meetup here.
I love to have time to get to know some of you. Although, this might be your first meetup after the meeting or the presentations today. We've got our meetup URL, the YouTube URL, so the presentation tonight, we do keep the archive of all of the presentations from our meetings. And we have the captions associated with the videos and the trainings that we have here.
And it's a great resource to go back over the last few years of what the meetup has had a lot of great speakers. I want to quickl thank our sponsors. We have Level Access who has been for pretty much our existence providing sponsorship for the communication access realtime translations that we have. I want to thank Internet Society New York Chapter. This is Joly MacFie in the back, thanks to the New York Public Library, the Andrew High school library. White Coat Captioning has been the provider for the text translations we see up on the screen. Equal Entry is my company and we focus on getting this event organized and happening on time. And our new sponsor tonight is Ira. So in our quest to continue to innovate and do new things with technology, we're very excited to begin working with this idea of
Ira, which is a new technology to provide audio descriptions live in a real environment. So just beginning to explore this concept. We have Marty Watts here tonight, who will be available to talk after the meetup. And we're also planning to have a meetup after this event to get more indepth on this technology. But it's very exciting technology. An we appreciate the sponsorship, basically, at this event. Someone who has signed up for the service to receive audio descriptions of the visual information occurring, we'll be able to get the audio descriptions for free at our events. We appreciate that.
Last piece before our presentation, so we haven't met in a while, but it's super exciting, here in New York City, the Access Plus Ability exhibit at the Cooper Hewitt Museum. We're very interested, hopefully, in technology and how technology and design can change people's lives. And so, this exhibit is all of this really interesting technology for and designs for providing solutions for people with disabilities. I believe it's going to be in New York City until September of this year. So you have a good amount of time to make it out there.
But I figured, I went this weekend, blew me away. Awesome technologies and stuff to learn from. It's a great exhibit. I wanted to share that with the group. If you haven't heard about it it's Access Plus Ability.
I'm going to be handing it over to the presenters for the evening. The National Center for Accessible Media and the WGBH. They have been pioneers in working this space, at least from when I started, working accessibility, I remember using their software 15 years ago, Mag Pi, that was a tool I used back then. Very excited tonight to learn about their new software and learn about new features, new ways to make media accessible.
With that, I'm going to hand it over to Donna, one of our guests tonight.
Thank you. Good evening, everyone. Can you hear me okay? My name is Donna Danieleski, I work at WGBH educational foundation.
I'm so thankful for the opportunity to present here tonight. Thank you, Thomas, and everyone for helping to organize this event. I'm here with my team today and we'll be talking a little bit about our caption and description editing tool. And providing a bit of training for you, the software, as you will hear, many times today is free. Check it out on our website, download it, play around with it.
There is a support group online, as well, for you to ask any questions there. And we're just thrilled. We're thrilled at the opportunity to be able to create it for you and to share it with you. And I don't want to go too far since we got a bit of a late start. I'm going to be quiet and introduce you to my two colleagues. I have Geoff Freed and Brian Gould. I'm not sure who is speaking first. Geoff. Thank you very much.
Don't clap, yet, until I deliver the goods. My apologies to people standing behind me, I'll be working on a computer sideways so at some points, my back will be to you. I'm Geoff freed here with Brian Gould. And we'll be showing you at a high level how our new captioning and description tool called Cadet works. As Donna said, it's free. And there are a lot of good things about it. For those who aren't familiar with the work, I want to give you an overview of who we are. NCAM was born in 1991. But our history goes back to 1971, which was when we funded the caption center and invented captions for broadcast TV. Here tonight and in TV and movies is rooted in Boston at WGBH. We have a very long history.
In 1990, WGBH launched Descriptive Video Service or DBS. And those of you familiar with audio descriptions, WGBH was a pioneer in audio description for broadcast television. And so, you'll be seeing Brian talk a lot about descriptions and the description of cadet. Mag Pi was the original one we released around 2000. It could be used to write captions for online purposes. But it fell behind the times. With funding we got from the Massachusetts Attorney General's office, we redeveloped MAGpie into something calledCadet. A browserbased caption editor.
It runs on the Mac, runs on Windows, I'm told in Linux. People say it works fine there. Because it runs in a browser, it uses HTML. We can write the software itself to take advantage of the accessibility stuff. If you are a keyboard user, if you're a screen reader user or a user of some assistive technology, you will be able to use MAGpie, as well, I'm sorry, CADET. I make this mission take all the time. I was involved in MAGpie development 20 years ago, and it's an old habit. When I say MAGpie, Brian will say CADET.
CADET is used for writing what we call preproduced or offline captions. A tool to write captions for programming that is already produced. This is not for live captioning. That takes a whole different skillset on the part of a captioner and a whole different technology set, too.
MAGpie can be used for preproduced material. I'm going to walk yo through the captioning half. And Brian's going to walk you through the description half. And I encourage you all to ask questions as we go along rather than save them up until the end. We've got about 50 minutes. So if you've got questions, raise your hand.
Otherwise, I'm just like an out of control train, which is not the thing to say, given what has happened. What we have on the screen now is a completed captioning project. And I mean, this is a project that contains a video that I want to caption. And it contains captions that I wrote. MAGpie is made up of three components, okay? Oh, geez. Is that why you shook your head?
The captioner needs to
Yeah, I'm sorry, I apologize to the captioner. Over here, we have a media player. This is where the video is going to play in CADET. Over here, we have the editor. And this is where you will type captions or write captions, and do some other things like time the captions. And then, down below, which is covered right now by the captioning window is a little status area where you can see error messages or progress messages and things like that.
Yes? Question? She can't follow the instructions on the website. In the okay so you're not able to okay. Why don't we talk afterward? I know the guy that wrote the manual. So I'll go through that with you. He's a nice guy.
So what we have here in our Finished Caption Project is the text, over here, the captions and the video. And what I'm going to do now is I'm just going to show you what the finished project looks like. I'm going to show you the captions playing in the video player. And then, I'll show you how it's done. We'll run this quickly.
You know how frustrating it is to help your kids. A visit to the doctor's office can be a challenge.
No, I don't want you to take my temperature. No.
I wish I'd feel better.
Sniffle, sneeze, no antibiotics, please.
Pause it right there. Okay. So on the screen, what we saw were captions playing over the video. And those captions were the same text that we can see in the caption editor. What I'll do now is I'll erase all of the work and then we'll go through the steps that are involved in creating captions. And as we go along, ask questions.
CADET I caught myself. CADET can caption videos of just about a format. So if you're a video producer or have a library of videos that need to be captioned, MP4 is always the happiest format, but there are other formats that CADET can handle, as well. Just about any format that will play in your browser. If you use Safari, FireFox or Chrome on Mac or Windows, CADET will handle whatever format that is. Most people use MP4 these days.
The first thing you need to do in order to write captions is open a video. So we will add a video to CADET. And here's the video. The same one we just looked at. There we go. Okay. No captions okay. That's funny, we see captions here. All right. Well, let's let's quit and restart. I don't know why that happened. But, okay, let's try it, again. Of course, it worked fine in the hotel room.
Okay. Try it, again.
Question is if you're selecting a video, can you open it as a file on your computer? Open it from YouTube, all of the above?
Good question. If you are selecting a video to caption, the video should live on your computer. It should be local. Which is, actually, a nice feature about MAGpie is you don't need to be connected to the internet in order to use it. Kind of a nice privacy thing for one thing if you're in an academic setting. That can be important. Also, you can download a bunch of videos on the computer and take a long flight and caption them in the airplane, if you need to, for example. Or any offline situation.
Once you write the captions, you can then send the caption file to YouTube or video or any other hosting service or use it in your own player. Now we have a video loaded into CADET. Okay. There we go. And so, now we need to write some captions. Okay?
There are several ways to enter caption text into CADET. The most obvious way is that you can type directly into the editor. And the way you do that is by playing the video for a few seconds, stopping, typing. Playing, stopping, typing. One nice feature about CADET, you can play it at slower and faster speeds so you can slow it down.
Which is probably not the thing to do at 6:00 at night.
You know how frustrating it is to help your kids when they're sick.
We could type that. And so forth. Okay. After you type one block of captions, hit the enter key twice, a new row will appear on the screen, and you can type. And then, you can go, enter, enter, and type some more and type some more. This is a perfectly valid way to enter captions into MAGpie but might be a little tedious, might be slow. CADET, sorry, and MAGpie, too, the same thing. But there are faster ways to do it, okay? A tagser way, let's clear out what we did. Get rid of everything. A faster way is actually to transcribe the audio by using text edit, or even Word as long as you save it as a text file. And I did that ahead of time using an application called BD edit, which is a plain text
editor. You can type your entire script into this text editor and if you put an extra blank row between each block of captions, CADET will understand that to mean each one of these little blocks of text will go into one of those cells in the editor. If you insert these extra rows, double carriage returns between each caption, when you open the file when you import that file into CADET, it looks like this, which is a fully typed out caption project. So all of the caption text for this video was transcribed outside of CADET and imported. This is a much faster way to do it.
This also means that if you use a speech to text application, you can transcribe, perhaps, by respeaking the audio into DRAGON, clean up the transcript and import it into MAGpie, you can use YouTube if you like to use YouTube to create auto captions but not just use auto captions in the natural state because they're not very accurate, he said bitterly. You can download the transcript from YouTube, clean it up in MAGpie CADET and send it back to YouTube. Okay?
I know, I'm having a rough night. Okay. So now, we have all of the captioning text in the editor. We now have the text, but we have no way to display the text with the video. That's because captions require timing information that instruct the captions when to appear on the screen and when to disappear.
And CADET will do that for you, too. It's we like to think a straightforward operation. The way that it works is you play the video. And you listen to the video. And when you hear the words spoken that correspond to the first caption or in this case, the first sound effect which is a woman sneezing, you press a key, which in CADET is control, hold down the control key and press the comma, and you will see a time code at that instant enter the "start" column, which represents when the caption should start its display.
And then, CADET will the little green focus will move down t the next caption and sort of arm itself for the next key press. And when you press control comma, it'll move down, again. And so forth. And I'll show you how that works now.
I'm going to have to put down the mic because my hand is not actually, it is big enough. But I don't want to take the chance. I'm going to put down the mic and speak loudly as I do this. I'm going to play the video and watch the time codes in the "start" column. We'll go through this video about one minute long.
Oh, wait, should pick up the speed. Okay. Let's go back. There we go.
You know how frustrating it is trying to help your kids when they're sick.
I don't feel so good.
A visit to the doctor's office can be a challenge.
No, I don't want you to take my temperature, no.
I wish it would feel better. My ear.
Snort, sniffle, sneeze. No, antibiotics, please.
As parents, you want your child to feel better as fast as possible. It's tempting to think that antibiotics are the answer when your child is sick, but that may not always be the case. You don't look so good. A lot of illnesses can be caused by viruses and antibiotics don't work on viruses. In fact antibiotics are strong medicines that could have serious side effects.
Stop there. That's about one minute of captioning, okay. And because I, of course, have seen this video a few times, I'm able to anticipate when to press the timing command. I always advise people, if you're going to be captioning a video for the first time, watch it ahead of time, or at least watch five minutes of it so that when you actually start timing, you'll have a feel for when you need to anticipate when the words are spoken.
Let's look at what we did. Okay. Now we have in the caption editor, we have a time code in every cell that represents the start time, which is when each caption is supposed to appear. And some of the captions have a time code in the end column, which means, the caption will erase because no words are being spoken.
So we say that's erasing the display. And then, the following caption has a time code indicating when it should appear when t person, the narrater is speaking, again. Now, we've done that. Let's go back and review a little piece of this. As you all know, when captioning, it's easy to make mistakes and you should review your work and make corrections. Let's do that now. I'll play the video. I know that I made a couple of mistakes. I'll stop and correct them and show you how you can correct some errors, and we'll move on. Here we go.
Okay. That first caption, I think she sneezed a little bit before the caption appeared. I'm kind of persnickety about timing. I'm going to use the keyboard to have a nice, precise timing point.
And now that woman sneezing just beginning to appear on the screen, I'm going to change the time code by simply pressing the timing command. On the keyboard, the keyboard command for CADET to back up a little at a time is "control G." It's the same on Mac and Windows. Now, let's play forward and see what's next. That was a little bit late, too. Let's back it up a little. And retime it. I timed the wrong caption, so let's back her up, again. Correct my mistakes.
G, move it down.
You know
Okay. There we go. Okay.
Know how frustrating it is trying to help your kids when they're sick.
I don't feel so good.
That looks pretty good. That's good.
can be a challenge.
No, I don't want you to take my temperature, no.
I wish it would feel better, my ear.
Okay. Now, there will be one thing here. I'll show you how to change.
Snort, sniffle, sneeze.
The captions are at the bottom of the screen and covering something, a visual element, covering up words. We can move CADET to move captions to the top of the screen. And the way you do that is simply highlight the caption rows that you want to move. We have two rows, and you can choose a style, for the top, and the captions will go to the top. You can move them to the top and the editor, there's a little carrot that points up to the row number that indicates these are at the top of the screen.
Okay. So in the interest of time, I'll move on to the next thing. A couple of things to do when correcting your work. If you have a fiveminute video, you should expect to spend half an hour on it, writing the captions, fixing the captions, fixing the work and doing something with the captions. Now you've created a caption file, timed it, edited it, checked your work, what are you going to do? Chances are really high you're going to do one of two things, send those captions to a video hosting service like YouTube or Vimeo or embed them in a video on your web page. You need to do a process called export. And export turns them into a format that a video player can understand.
A couple of very common formats that some of you might know about are called web VTT, another one is called SRT, another one is TTML. And there are others, too. If you use YouTube or Vimeo, Web VTT seems to be the one they like, also SRT. Browsers, if you're embedding a video into the browser, they all like Web VTT, some of them like TTML. So let's export this file in one of the formats. And we'll export it as web VTT. Okay? And CADET's telling me I've done it once, but I'll tell it to overwrite. And now, when it extorts, gives me a message. And now, if let's do it this way. Okay. If you have a video that you wan to I'm sorry I'm covering up the captioning window. I'm about t uncover it. I've lost the caption window. Uhoh, sorry. Hang on.
Is it in the same browser? Oh, I'm sorry. My apologies. Come on down here. Yes, I'm very sorry. Okay.
Okay. Apologies for that. Here we go. All right. So here is a video embedded into a web page. And these are the captions that we just wrote.
You know how frustrating it is to help your kids when they're sick.
I don't feel so good.
A visit to the doctor's office can be a challenge.
No, I don't want
I know you don't want it. These are the same captions that we just wrote embedded in a simple web page. These would also be the captions you upload to YouTube or any other service. That caption file is what you would reference, which is a bigger topic than we have for tonight.
A little bit of a rushed fashion, that's how CADET works. Yes?
Can the YouTube video player support the captions?
A good question. Yes, a very good question. We have a rule at NCAM that says just because you can do it in CADET doesn't mean it'll show up as you expect in any one of these players. And that is because different browsers and different players support all or some of the captioning specifications, like SRT and web VTT and stuff.
And so, the answer with YouTube is, no. And the answer to Vimeo is I'm pretty sure no, as well. One thing nice about the YouTube captions, the user can drag them. Click and drag them where they want out of the way.
So when you use a special feature, like, moving the captions or coloring the captions in CADET, you should always test it in the target player to make sure it's supported.
Okay.
One more. Yeah.
open for everyone to see? Or can you turn them off or on?
When you upload the captions to YouTube, are they open? Which means, they can't be turned off. Or are they closed? They are closed. And there's if the video has captions, there's a little button that surfaces in the player. And you press the button and you choose the captions you want to see. Yeah.
Yeah. And there is a way to disable the automatic captions, as well, which is available on the YouTube website, which you should do if you're going to be putting up real captions. In the interest of time, I'm going to turn things over to Brian now who is going to talk about audio descriptions.
Hi, I'm Brian Gould a colleague of Geoff and Donna's and I'm going to talk about audio description. Audio description was pioneered, actually I'm going to put my pieces paper somewhere where I can see them. It was pioneered, also, at WGBH like captions in the early 1990s. And audio description is a similar service, except almost the reverse of captions, in that, audio description is obviously an audio track that's added to the sound track of a TV show or movie and a narrater describes the visual elements of the scene. And it's woven seamlessly, hopefully, in the natural pauses in the dialogue. And if anyone's not familiar with description, I'll play just a little snippet of it. And then, I'll show you how you can use CADET to write description scripts.
And this isn't going to work because it's not my computer. Whoops. That's okay, because when I'm done writing the script, I'm going to show you a described video. What I have here is the CADET Editor. And I'm going to move my video to a predetermined point.
In a video. So the set up is similar to captions. Although, since I've moved project type from captions to descriptions, I should actually move up here, I've gained a couple more tools. I still have start times and end times because you need to know when the description needs to be spoken.
Now, I have something a new column called an audio queue. It's very important to know that with CADET, you're not creating an audio track. The output of CADET is a script that's intended to be either recorded by a human narrater or read out by text to speech and the way that process would work in embedded that into a video is a presentation for another day. But let's put it at you're writing a description script that will then be recorded in a separate process and then added to the video. So if you have a narrater, usually description scripts are recorded in realtime. You play the video and then, as a pause comes up, the narrater would read the description and the audio queue helps them know when they should start reading.
Also, narraters will watch the time code to know when something's coming up. And also, when it ends. That's why you have those. The caption column has turned into the description column. And, of course, it doesn't matter how much you write. You need to write as much description as you have time for if there's a long pause, you're going to write not very much in three to five seconds, but if there's a long pause, 90 seconds, two minutes, you'll write a lot of description and you'll fill up this entire you can fill pages and pages, it's fine.
It's different from captions. Captions, you want to keep to a certain number of words or maybe two lines so that it can be read. It's a different case where you're going to be hearing it for captions. You also, at the top, have a really vital tool for anybody writing descriptions. And that's the times window. It gives you it says description and pause times. And let's dive right into creating a new description. And I'll show you the way that this works. I'm going to go ahead and hit play. Actually, I know I'm set up for the beginning of a pause.
I'm going to go ahead and put in the end time for this pause. And, again, the process of writing descriptions in captions is quite different. For captions, you can write a transcript and time it. For descriptions, you can't do that because you could write a description of the scene and find out, there's only five seconds of silence in the entire scene. You have to kind of map out how much time, how much pause, so to speak, there is in a scene so that you know how you're going to approach the description. Can you give a lot of information setting up early? Or is it very dialogue heavy early in the scene and you need to weave in little bits of description as the scene goes on?
What information are you going to be able to provide when? So as a describer, you need to time out maybe a minute, maybe longer so that you have sort of that map to know what you're going to write when. I'll go ahead and play. And as soon as I hear dialogue starting up, I'm going to stop it, rewind it a little bit and create my out time. Or my end time.
Okay. That started dialogue. I'm going to jump back using a keyboard command and I'm going to create my out time. I know, if we look up at the pause times up there, I have 4.08 seconds. So as an experienced describer, I know what I can write in there. I started writing descriptions at WGBH in 1996. I've been doing it for a long, long time. However, I have my notes, I know what I'm going to write. And we've already set up we're in a yurt and I'm cheating because I've watched this probably 500 times and I know what's about to happen. In a yurt the shaman studies the spear.
The narrater may be looking at time code, but it's also helpful to have an audio or visual cue. And I know that the cue here is cut to black. I'll go ahead and double carriage return and I'm ready to map out the next pause. My next opportunity to provide a description.
Well, we're getting interested here. Okay. So we'll go ahead and put in my end time. And now, we'll listen for when dialogue starts up, again.
You're a fool for
Now, I'll back up a little bit when he just ability starts talking. And, again, I have 2.28 seconds, I can't say a lot. Again, normally, a describer would do this process for one or two or three minutes far into the scene. I know exactly what's happening and I know what information I need to give because we're about to hear a sound effect of him pouring soup into a bowl. And if you can't see it, it's very odd. You don't know what's happening, what is he doing? What is he pouring? So he ladles soup into a bowl.
And I've spelled that wrong, but we have spellcheck. A little red squiggly in CADET. It's not MAGpie, it's CADET. I know I need to tell my narrater he just said "innocent blood." Okay. I need to show you the best part for writing a description. If we look up at the times window, we have our pause time is 2.28 seconds, and now the window has turned red and I have a little musical note, which means, look up here, there's a note. CADET automatically estimates the reading time for a narrater and tells me I've written 2.8 seconds of description. Okay. If you find that your narraters speak faster or slower, that reading speed is adjustable in CADET.
I know this works well. I know through experience, that's not this is not a vital error We're not off by too much. I could go back, squeeze a few more tenths of a second. And because they show up on the screen just like captions, I can go ahead and play it and read it myself and say, does this really fit or not?
Now, if I was off by a second or two seconds or five seconds, I would know I have to come back and edit a lot. And it's very helpful to know how much time you have available to you. So I will back up. What am I doing wrong? There we go. And let's play this.
He ladles hot soup into a bowl.
You're a fool for traveling alone.
Even added an adjective. That was good. Let's write a couple more. And then, I will show you something neat.
Completely unprepared? You're lucky your blood is still flowing.
Okay. You're lucky your blood's still flowing, that's the beginning of what I know to be a pause. Oh, my description is, he hands her the bowl. And how did I do there? Oh, I forgot. I look up to my pause time to see how I did and forgot to put the out time.
So I need to back up. Which is very easy to scrub back and forth, and you can do that by very tiny increments, sort of medium or large increments. I don't remember what we have it set here. Maybe 1/2 second, second, second and a half just by different key command. Also, those are completely adjustable. So you can set your scrub parameters however, whatever works for you.
You're lucky your blood's still flowing.
Thank you.
So we want to get out before she says "thank you." That should be right. And really, if you're off by a second or more, that's when you're going to be in trouble. And if you have 13 seconds, 25 seconds, 50 seconds of pause and you're writing, there's no way to estimate that in your mind or even in your reading, you're going to naturally, sometimes, speak too fast.
So CADET gives you this tool that, actually, is a check against someone's natural tendencies to keep all of the beautiful words that you've written. If you need to cut it down, you need to cut it down because you want that delivery to be as consistent and as flat as possible. You don't want your narrater speeding up or slowing down. You want it to be concise and consistent.
And, again, this tool helps you write a script that can be recorded. The tool is not going to write descriptions for you. It's also not going to make suggestions about what's the context of this film I'm writing? Or this show, am I writing educational programming for early learners? Or am I doing sort of an adult drama or an action film where your vocabulary, your sentence structure, your sentence length, your assumptions about what your audience has experience with are all going to be different.
So the content of description, again, is really focused on the context of the video you're working on. CADET's not going to help you with that. It's going to help you time things out and then, provide a script that you can record later. Let's go ahead and go into review mode. I'm going to back up, actually, to the beginning of this film. What we've been doing in edit mode is the let me think about this as you create, you' creating new descriptions. And then, you're playing the video. And sort of the you're moving the video to where you want to be. In this case, when we switch to review mode, the big change is that when I hit play, my cursor inside of the editing window is going to follow the film.
So the reason for that is, maybe 25 caption descriptions down, I find an error. I don't want to have to hunt for it. CADET's going to have you right exactly in the column or cell you need to be to make your edit. Okay? It also does something else.
We saw it displays the descriptions just like captions so you can read them without having to look inside the description field. You can read them right off the screen. There's also a really neat little feature that's so helpful for editing. And I'm going to go ahead and play it right now. But it didn't work.
I didn't have the cursor at the beginning. Here we are.
In animation snowcapped mountains. This film was supported The Netherlands film fund. (Inaudible).
So what the heck is happening? So CADET, this is the one feature you need to be connected to the internet for. It is actually reaching out to an open source free TTS engine and reading the captions out loud so that you, as the description editor can experience what the final product will be like. And that is really, really helpful because then you don't have to be reading, you can sit back and watch it as an audience member and see if your sentence structure is appropriate. If you use good vocabulary, if the if the descriptions really do fit in the spa you want.
And you can adjust. And it's a different experience reading yourself or hearing it as the final product is meant to be. And I don't know any other tools, certainly no other free ones, that have this feature. And I've found it incredibly helpful . Let's see, anything else? Yes.
So, we've written all of our descriptions, timed them out, edited them and reviewed them, we know they're perfect. So we can go ahead and export. And we have a description export. You can export into any of these formats. You wouldn't export to SRT or web VTT, but you might export as a plain text file. But really, you want to export into CADET's description format.
And this is why. Because you get a nice HTML script that gives all of the information in a format that's very, very helpful if you're using an audio engineer or producer. And also, for the narrater. You know which caption event or description event you're in. You have the "in time" to look at, here we have "fade in." I'm going to scroll down a little bit. Once we have some dialogue, you can see your audio cues right here.
The description, that's going to be red, is in very large, bold font. And then, the narrater may want to look at the same time of the time code and know when to stop when they're going to need stop talking. And that can be either handed to someone on a laptop or on a tablet or printed out in this way. And this is a really simple, but very clean and useful script format.
Going to click back into CADET. And that was sort of really quick. But I think I'm ready to answer any questions if you have them about writing description with CADET. Yeah?
(Speaker off mic).
I'll repeat, I think.
I have a question about the description captions. Does CADET only support for English?
So the question is does CADET in captions support multiple languages? And the answer is, yes. And how many? Or are there any that it doesn't?
You can write captions you write captions in CADET in any encoded language that you can type into your computer. Let's put it that way. You can do the same for descriptions, and you can certainly generate a script, like, what Brian showed you and printed out. What the text to speech engine does with that, that's a good question. We've never played with that. My assumption is that it's probably going to mangle it. But I don't know. I don't know.
The real answer to the question is, yes, you can write about anything you want into the editor. Okay?
Any other questions? Yes, right here.
Are you able to export, not just the description
So the question is, are you exporting a description, only the script or also that TTS voice along with it? And the answer is, you're not. We haven't moved forward to try to do that. The text to speech engine being used is great for the editing process.
I don't know how great it would be to listen to. It spells some things out when you wouldn't expect it to. You would have to do some fudging with it. Yeah. Also, technically, there's some challenges. We can talk about that.
You could. If you wanted to export raw text file, feed that through another text file like text aloud or something else that maybe is more sophisticated. And you can create an audio track that way. With, perhaps, nicer voices and edit that into your video.
Maybe you claim all of the technicalities.
Why don't you talk about what happens with the script?
The script, then, needs to be recorded. Most typically, the way it works is that a human narrater reads the script and that's recorded. And then, there's a post production process where that is integrated into the original soundtrack. Fortunately, with most video platforms, YouTube and others, there's no way to have, like captions, a toggle on and off. There's no description on and off.
You'll have to offer an open described video or two versions of the video, one that's described and one that's not. That's not the fault of CADET obviously, that's the fault of HTML5 not allowing you to do that or, really, technology.
Joly: Following up on that. So there are no standard formats, schemas, or something?
Yes, mostly correct. Yes. There isn't any formalized format for, say, exporting the description script that Brian wrote into an xml format that all text to speech engines will understand. Unlike captions where you've got a number of standardized caption formats.
Joly: Have you talked to the W3C about that?
Yeah. Sure. Yes yes, we've participated in a number of working groups around various kinds of markup languages. There used to be a style language, called audio style sheets which I don't even know if that's been used in decades.
But the idea behind audio style sheets is that you would use styles like CSS for text. You would use styles for audio. And as far as I know, audio style sheets went nowhere. So, am I right? Yeah, someone's giggling in the background. So there has long been a need for something like what you were talking about but there really has been no standardized approach.
Which is why one of the suggestions we make is you use a raw text file and feed that into it. That's one approach. But browser and audio and visual players can't do that text to speech for you. As far as I know they are custom yes there are custom players that do do that.
And there are other approaches where you can actually feed invisible descriptions to a region that a screen reader can read aloud on the fly. That's another approach another presentation for another night.
So you can have prewritten time of descriptions that are fed and a screen reader would read while the video was playing. It's a custom solution that's not widely available. We have another question.
The question is do you have to see the do you have to see description text? Or can it be hidden? Do you get my question?
Yes. I actually don't know the question.
The question is, yes, can you turn the descriptions off in the caption editor?
But still have it?
The answer is, yes. Like that?
Yes.
The audio would still be
Yes, the audio would still play. If the captions and editor are distracting, you can turn them yes
Oh, from the
Separate files, separate data.
Okay. So the real question is when would you ever have both up?
You would never have both up. And the reason you would never have both captions and descriptions playing in the editor because this is an editor and not a player, right? So you're either editing captions or editing and creating descriptions there's probably times as a player you would want have the option of either descriptions or captions or both or none, but in the editor, we're limited to only having one or the other. And the technical reason for that is because captions and descriptions are just time text file that's being created and displayed on the fly as they're being created.
So CADET doesn't know the difference. It's just displaying the time text that you're creating. It's the mode whether you're in description or captions. The way it displays or whether it reads it out loud. I hope that wasn't too confusing.
Then, you upload it to YouTube?
So you would have to create the sound so, the issue is that there's not yet a seamless, elegant way to write a description script and then have it described to video. Nobody does this. CADET enables you to have a script that's timed and that you can record a narrater reading it, and then you have to marry that description track to the video. And that would be uploaded to YouTube or Vimeo or your video platform is.
The question is, because people listen to screen readers so much faster than natural human speech, you could write a lot more description. We have done some research and testing with this. Mostly as the description voice. And what we found as a result of that testing was that with shorter howto or academic videos, the text to speech voice was fine and the better the and we did this 5, 7 years ago, and text to speech voices were getting good. Now, they're quite excellent.
But when you are watching a film or any video for entertainment, the overwhelming consensus was that text to speech didn't cut it. Human narrater was much more pleasurable. It didn't make mistakes. It didn't stumble over weird words. And those are all of the those are some of the issues that you have with text to speech.
But it's also not been done in any widespread way. So, once it is, probably people will be able to improve it.
(Speaker off mic).
Basically, this space, we have really hard cutoffs. So we appreciate everyone coming out today. We can continue the conversations and discussions outside. But the library has a hard stop. So we appreciate, as always, everyone coming. It's a great talk tonight. Definitely, going to be using CADET myself. So thank you, guys. And we'll have our next meetup on March 6th, the first Tuesday of the month. Stay tuned to the Twitter and Meetup page.
We'll be announcing that shortly. Thanks. [ Applause ]