3Play Way: ADA Title II Compliance [TRANSCRIPT]
LILY BOND: Thank you for joining today’s session, “The 3Play Way– ADA Title II Compliance.” To introduce myself, I’m Lily Bond. I’m the chief growth officer at 3Play Media, where I’ve been for almost 11 years. I am a white woman in my mid 30s with brown hair and glasses, wearing a gray sweater, and I will hand it off to Jessie.
JESSIE ZIONTS: Hi, everyone! My name is Jessie Zionts. I’m a senior customer success manager here at 3Play, working exclusively with customers in the higher education space. I am a white woman in my late 20s with brown hair, wearing a beige sweater, and my pronouns are she/her.
I work with tons of people in the same, if not very similar, position to those of you who have joined today’s webinar. I ultimately just help folks find success with 3Play services. And it’s very possible a handful of you joining today are one of my customers here at 3Play, so if you’re here, thanks for joining. We look forward to discussing ADA Title II compliance with you all and learning how 3Play can help solve common roadblocks in this space.
LILY BOND: Great. As a quick agenda, we’re going to start by covering an overview of ADA Title II compliance, just to make sure everyone’s on the same page with the regulations, cover a lot of the things that we’ve found in conversations with our customers and the common themes in how to build compliance strategies, and then go through the solutions that we’ve developed as a result of those conversations to really help universities tackle compliance through the 2026 and 2027 deadlines and moving forward. And of course, we will leave time for live Q&A.
Just to make sure everyone’s on the same page with the rules and regulations of the ADA, what we’re talking about today is Title II and the updates that went into effect over the summer of– it’s already 2025– over the summer of 2024. These updates modernize Title II of the ADA to apply to the web, and requires WCAG 2.1 AA compliance– that’s the Web Content Accessibility Guidelines– with compliance phasing in in April of 2026 for jurisdictions over 30,000 people and April 2027 for jurisdictions under 30,000 people.
Who does this apply to? ADA Title II covers public entities. This includes public colleges and universities, as well as K-12, state and local governments, and public services. And we’re going to be covering specifically how the ADA applies to video accessibility and how we can help universities make online course content and video content accessible. And the requirements for video under Title II are closed captioning, transcripts, and audio description for recorded content and live captions for any real-time content. And I will hand it off to Jessie to cover some of the conversations that we had with our customers.
JESSIE ZIONTS: Awesome. So we proactively reached out to a handful of our education customers here at 3Play and asked if they’d be open to having a research conversation with us so we can learn more about how they’re approaching and interpreting ADA Title II compliance. We had 31 awesome conversations, where we learned tons of amazing insights on how schools are planning to enhance their digital accessibility efforts, and these conversations actually helped us shape our new product offerings to exactly what those in the edu space are looking for.
We asked all different types of questions in an effort to really understand how each school and university is solving each piece of this complex puzzle. We dove into general planning questions, questions like, who is driving compliance, and how are you planning and prioritizing? We got into technical questions like, what is your timeline for compliance, and is this budgeted? If not, are you building a budget?
And lastly, we brainstormed around potential solutions. We talked through what solutions are you considering? Which services is your content missing– closed captioning, audio description? And what approaches are you looking at? Have you considered AI involvement?
And here’s what we learned about the market pressures and solution concerns that you’re all universally facing. Compliance drivers– what we’ve learned is that the drivers are coming from multiple different division heads saying, we have this problem as an institution, and we need to come together to figure out what we can do about it. This could come from the accessibility, IT, or finance team. And from there, we’re seeing it’s most common that subcommittees are created in an effort to delegate work and be the most efficient.
Audit and prioritization of content– this is a big one. People don’t know where to get started. Most schools and universities have years and years of content posted in every which way direction, and it’s an overwhelming thought to try and make this many years of content accessible. Schools are looking for clarity on the vague regulation language and, of course, determining a prioritization path. With this many years of content, schools are needing to determine which digital media content takes precedence.
Lastly, faculty adoption and centralization, two things that can notoriously be very challenging– schools are seeing a lack of faculty buy-in, which could feel like a result from accessibility training gaps, and faculty often prioritize speed and ease of use over centralized systems, leading to fragmented accessibility practices. Different departments have different budgets, resulting in a question mark on where these funds would come from.
Now, even if everything was centralized and everyone was aligned, there are still major concerns when it comes to finding the right solution. Schools are wondering how they can effectively and efficiently produce closed captions and audio descriptions at this scale. Cost has come up quite a bit as a concern– and I’m sure quite a few of you are nodding your head in the audience– especially with audio description, since the service has traditionally been such a high cost.
Accuracy and compliance, with concerns around the use of AI and how heavily humans should be involved in editing if AI is being used. However, high accuracy in captions in compliance with legal standards is always top of concern. And concerns around turnaround times– services need to align with academic schedules and demands.
Workflows and features, this is a big one as well. You all already have designated video platforms and LMS platforms, amongst many other things you’re managing, so digital accessibility solutions need to effortlessly fit into these workflows. And I’d love for you guys to chime in in the comments if you agree with this sentiment of what we’ve learned and the concerns. Or if you feel like we missed anything, we would love to learn.
So after speaking with 31 different schools and folks in positions just like those of you on today’s webinar, we have identified three common approaches to tackling these new requirements. Strategy will be key in order to prioritize different types of content. There needs to be a clear way to determine which content is tackled first and with what services. Colleges and universities are looking for partners with expertise to act as an advisor while they navigate these new regulations. And with tight budgets and turnaround constraints, a combination of AI and human involvement is necessary for tackling these large amounts of content.
LILY BOND: We have launched three new product sets for universities. These are really targeted at– when you think about prioritization, you have ongoing content, you have public-facing content, and you also have, in many cases, large libraries of courses that now fall under this Title II compliance. And these backlogs are very high volume, and we believe that these solutions are really best suited to tackling those backlogs.
So the first product here is AI-Enabled Audio Description. We’re really, really excited to launch this in market. It is the first product of its kind in market. We received a patent for it last fall, which is really exciting. And this is an AI-Enabled Audio Description service with upgrade paths to human-reviewed audio description, so really unique in market, being able to support both the AI audio description and the upgrade paths to human review where necessary.
The second product here is Predicted Caption Accuracy. For backlog content, you may be considering ASR, or Automatic Speech Recognition, to tackle the captioning portion. And we are able to predict the accuracy rates of that ASR so you can set thresholds for quality across all of your backlog files, allowing you to mitigate risk at scale. And then, finally, one of the services that we’ve heard a lot of demand for, we’ve now integrated into our live captioning offering, and this is Live Summarization. So for live or real-time accommodation, students can now choose between verbatim live captioning or AI summarization of a lecture.
I’m going to go through each of these products in a little more detail. To start, I’ll cover our AI-Enabled Audio Description service. This is an audio description service that supports both standard and extended descriptions. Standard description fits descriptions of visual information into the natural pauses in a video, whereas extended audio description will extend those pauses to make room for description if there is not sufficient space.
This service uses optimized AI to script the description and to voice the description. And as I said, we have an upgrade path to human review, or customers can use our in-platform audio description editor to edit that script yourself and reprocess the file for description.
How do we do this? This is really exciting. Our data science team has done an incredible job working on this product, and we’re using state-of-the-art AI software to analyze a full video. So we’re not breaking this video into two chunks and describing different pieces of it. We are describing one entire video and producing a description using best practices for quality description and nuanced feedback from our expert human describers.
I think this is really important. Description is hard. Really good description is really hard. And we have decades of experience creating description with in-house expert describers. And that team was deeply involved in this process, making sure that our AI results followed all of the best practices and as much of the nuance of human description as possible. And the results are really exciting.
As I said, this process is patented. And we believe that this is going to enable schools to make audio description widely available. It has been really hard to publish audio description at scale because it is so expensive historically. And the turnaround is also longer, so this is a really expensive and time-consuming process. And AI audio description is going to enable schools to really create more widely available description, as I said, with the escalation path so you can set a threshold for quality and make sure that you are having strong quality description across the board.
I am going to share a demo of the AI-enabled description right now. I am going to show two videos back to back. The first one is more of an interview or lecture-style video, and the second one is more of a campus tour or public-facing video. These are both Creative Commons videos. And I would love to hear feedback in the chat as we go through. And I’ll cover some of the things that I would call out about the description afterwards.
[VIDEO PLAYBACK]
– Text– how would you characterize online learning? Professor Ray Schroeder sits and gestures.
– When we begin talking about online learning programs, we really have to begin with the student. The success in online learning is building a relationship with that student in providing effective learning opportunities.
[END PLAYBACK]
[VIDEO PLAYBACK]
– That’s Berkeley.
– Berkeley.
– Students walk through Sather Gate. Several students face the camera. A football game. A lecture hall. A sepia-toned photo of people with shovels.
– Built by the people for the people. Founded–
[END PLAYBACK]
LILY BOND: Great. So I would love to hear people’s feedback in the chat. Some of the things that I would call out about this, in the interview, the most important thing to describe is the text on screen, and there’s very limited space to do that. So what is the most important text in that setting, and where can you fit it? So smart decisions about what’s critical. I think that description was really solid for the space available.
On the campus tour, I think everyone would note that the timing of the description was a little bit off. That said, it is the best practice to describe as much of the visual as possible where you have space to do so, and the description did achieve that goal. This is an example of a description where you might have this video on the home page of your website, and we would probably recommend upgrading this to human review so that you have more nuanced decisions about that timing.
But the description itself was really strong, accurate, and even called out things like the name of the gate accurately and specifically, so some exciting AI examples there to show how well the description can do for different types of content. I do see one quick question in the chat that I will answer, which is, is the description available in an American accent? Yes, we have multiple AI voices available, and customers can choose the voice by file or by project for any of their videos.
I’m going to move on and talk about our Predicted Caption Accuracy quickly. So this is really to predict the accuracy of speech recognition on a file-by-file basis. If you have a backlog of 10,000 courses, and you know speech recognition is on average 90% accurate, you might feel like that is good. You might feel like that’s bad. But that is an average. And the experience that we have as viewers is on a file-by-file basis.
So one file may have 95% accuracy, and another file that’s a complex physics lecture with a speaker with an accent might have 60% to 70% accuracy. And that file really needs human upgrade. And this tool is going to enable you to make those decisions without having to review every file. So we’re predicting the accuracy on a file-by-file basis, and then allowing you to set accuracy thresholds. So if you set 95% as your threshold for the backlog, you can automatically upgrade anything that falls below that to human review.
And this is using 15-plus years of corrected ASR data. We use this internally to make sure that we understand how difficult each file that comes through our system will be to edit. And this process is also patented.
And this is a tool that we would recommend using on backlogs, but we would not recommend for ongoing, net new content, accommodation requests, or public-facing content. We would always say that human-reviewed, 99%-plus accurate captions are necessary for compliance in those cases. But for schools trying to tackle very, very large backlogs with budget concerns, we think that this tool is a really good way for you to set some thresholds that you’re comfortable with and make sure that all of the files within that backlog are hitting that threshold.
And I have a quick visual of how this works. So there’s a diagram on screen, which I will describe. You would upload your backlog files. We would process all of those videos through ASR. And we would predict the word error rate for every single file. And then you can make decisions.
If you set your threshold at 98% accuracy, anything below that will go to the 3Play Media marketplace for our captioners to review. And we guarantee compliance with those videos, and then it would go to publish. Anything that falls above that threshold you set will go straight to publish without that review step. You can always self-edit those files, if that’s your preference as well. And I will hand it off to Jessie, who will cover Live Summarization.
JESSIE ZIONTS: So we’ve released a Live Summarization product that works as an extension of our existing live verbatim captioning tool. So moving forward, universities will have the option to add on this service to their existing 3Play accessibility tool belt at no additional cost. So what is this service? We will leverage AI to summarize lecture content in real time alongside the student’s verbatim transcript. We’ll see an example of this in the next slide.
And how does this work? We’ve developed our AI tools to use specific prompts in order to produce comprehensive and digestible short summaries based on what’s being spoken out loud. And why did we create this new service? Some students prefer this style of note-taking, paraphrase, meaning-for-meaning, summarized version of verbatim transcripts, and we wanted to be able to offer this as an option to these students, knowing that both the accommodation requirements and personal preferences are met in one convenient view.
Not only that, but this removes the need for DRC members to juggle multiple different vendors and processes. We understand that some universities have restrictions around summarization tools, but for those who don’t, we wanted to offer a chance to enhance your students’ live captioning experience with this new and exciting offering. And I’m very excited about this offering. I do a lot of live captioning work with my customers. I’d love to hear in the comments if this is something that you think your students would like and how you feel about it.
So shown here is a side-by-side example of our verbatim live transcript compared to our summarized live transcript. As you can see, this tool is taking longer, wordier sentences and breaking them down in just a few words. Our Live Summarization tool always includes an accessibility toolbar for your students to customize their viewing experience to their liking. And for those who have chosen to utilize this tool, students will now see a transcript option on the top left-hand side of this toolbar, and from there, they’ll be able to toggle between the verbatim transcript to the summarized transcript. They could also view both options side by side, as shown in this example, if they’d like.
LILY BOND: I wanted to go back to a compliance timeline before we close up and go to Q&A. One of the ways that we can help partner with schools is in thinking about what’s necessary in partnering with vendors to meet compliance. So we have some kind of best practices based on the conversations we had and the ways that we know contracting and onboarding typically work. Right now, schools should consider being in a solution-gathering phase. And I say that because in order to get these new tools and the compliance solutions you need not just for video accessibility, but for compliance in general, into the budget prior to the April 2026 deadline, they have to get in budgets this year.
So typically, schools are in the budgeting process for the 2025-2026 school year now through May, depending on your university, with fiscal starting in June or July. And during this process, it’s best to do some solution gathering, talk to lots of vendors, understand what solutions you want to use, and do a content audit so that you know how much content you have that needs compliance. Because that’s going to enable vendors to give you the best pricing and comprehensive view of what our partnership can look like for your full needs.
Then we recommend onboarding your chosen vendor over the summer so that you can start with net new content being in compliance with the fall semester. That way, you don’t have to go back and do this coming school year after the fact. We would recommend trying to get compliance in place for the fall, and then start processing your existing content or your backlogs once the semester starts and you’re out of the craziness of the beginning of the semester.
You have plenty of time to get through your backlogs, and working with a vendor over the course of several months to get that done will be important, and then again, remaining in compliance for the spring semester, feeling really good about your strategy to hit that 2026 deadline in April, assuming your jurisdiction is over 30,000 people. And then just a reminder to review what you’ve done with vendors and meet back with them in the ’26-’27 budgeting process so that you’re optimizing your contracts for the lessons you’ve learned over the course of the year and the content that you have a good sense of moving forward.
So our suggestions, based on the things that we’ve learned from you and the conversations that we’re having with schools, schools that have developed subcommittees and are auditing all of their video content are really well positioned to move towards compliance for 2026. And then we recommend having a clear prioritization plan for existing and new video, something like anything net new, anything on our public web page, anything with an accommodation request is getting human-reviewed accuracy for all accessibility solutions. And then for backlogs, we’re using a combination of AI and human. That type of approach is something that you should have an opinion on, as an organization, and work with vendors to make sure that you can move forward with a plan.
And with that, we will open it up for questions. We really appreciate all of the engagement throughout and look forward to the questions that are already coming in. You can always reach out to me or Jessie. Our emails are on the slide, [email protected] or J-Z– I’m going to spell it out– J-Z-I-O-N-T-S at 3playmedia.com.
And we would love to both share resources and meet with you, if you’re interested. There’s a QR code to download our ADA Title II checklist, which we’ll also put in the chat, and then a QR code to schedule a meeting to learn more. We’d love to talk about these solutions with you and hear what’s resonating and how we can help you with your compliance strategy. So that said, I will pull up some questions. And let’s get started.
The first question I have here is, how is the– this is in regards to AI audio description– does the product automatically add in the extra video space to fit the extended audio description? That’s a great question. So the way that we do extended audio description is we identify the best places to pause a video and add description. And then depending on how you’re publishing your video, we will produce various different outputs.
If you’re using our 3Play Access Player, which is a plugin, we will actually pause the video to make room for those descriptions without republishing a new version of the video. Or if you are posting back and publishing a new version of the video, we will automatically add in that extra space for you and give you a video output that includes the new duration of the video. So our product does kind of complete that space add where necessary, and depending on how you’re publishing is really the question for what the output looks like.
Next question, also about audio description, audio description would need to be turned on, correct? A video wouldn’t always have audio description built into it. This again depends on how you’re publishing. So if you are using a video player that supports an audio description track, you would have a toggle, just like you have a CC toggle to turn captions on and off. There’s usually like a headphones icon or a toggle for different audio tracks, where you can toggle between a described version and a non-described version. If your video player does not support audio description, then you would need to publish two different versions and link between them.
And there’s another question about whether it’s required or preference to always have the audio described version or have it available to toggle. The legal requirement does not state that it must always be on. I think a great parallel to think about is captioning. Captioning needs to be available to turn on or off. It doesn’t need to be turned on all the time. It’s user preference. It has to be available for them. Similarly, audio description– it has to be available for them, but the user can toggle it on or off.
Another question here, Jessie, I think maybe you could answer. I’m a 3Play customer. Are these solutions available to me now?
JESSIE ZIONTS: So the solutions are available. We would just encourage you to reach out to your account director or customer success manager to learn more about the solutions. We could set up a quick conversation just to make sure we’re aligned. And then we can either start testing files for you with the AI AD or get the Live Summarization tool enabled on the backend for your future events.
LILY BOND: Thanks, Jessie. Just reading through another question, it’s a long one. I’m going to read it out loud and internalize a little bit. “Our institute sits in a space where we, as a nonprofit, work across both academic clients and industry partners. We take this very seriously, but we are constantly struggling with what is the right solution to meet our needs while hitting our audience needs.
Is having AI captioning for live events and produced videos good enough? We have a streaming service that provides AI auto captioning that is pretty accurate, but we don’t have a staff to be able to go over every word in every single hour-long, two-hour-long, and eight-hour-long event.” That’s a great question.
AI captioning, or auto captioning ASR– there are lots of different ways to talk about it– is typically not considered compliant. In order to be compliant, you have to create an equal experience. And an equal experience industry standard is considered 99%-plus accurate. ASR is not there for a lot of reasons, so we would always recommend for a compliance solution, you are using human-reviewed captioning or live professional captioners with a human involved instead of ASR.
And then if you are using something like AI auto captioning, you would want to clean that up afterwards. When we’re talking about prioritization, and you’re considering something like a backlog versus something that is net new, something that is in a library versus something that is on the home page of your website, you, as an organization, should work with your legal counsel to understand what your risk tolerance is. And you can set rules like a threshold for ASR for lower priority content, as I discussed. But that would be our recommendation.
The next question here is, are your audio descriptions text-based that are machine read, or are they audio recordings? That’s a great question. So our description process, we create a written script, and then we use AI voices to voice over that script. So it’s not a human voice recording, but it is an AI voice. And the output is an audio track.
However, we do also provide the text description as an output so you can download, for example, a merged transcript that includes the transcript of the spoken word and the transcript of the description. So both of those are options, but the typical output for audio description is an audio track of the AI voice.
Next question is, how should schools prioritize which archived video content to make accessible first? So this is always up to the individual university, but a good rule of thumb would be to audit your video content and consider first what content has any accommodation requests against it and which content is most publicly visible.
The next would be to go to courses that recur every year, or every few semesters there’s a recurring course, and then go to the most viewed content. And then from there, I would work back by date. That would be my recommendation, based on what we’ve seen work well with universities in the past.
Couple more audio description questions here– with your patented tool, the AI is scripting the AD, yes? No human involved in any part of that AD creation. Great question. Our AI-Enabled Audio Description has AI scripting the description and voicing the description.
We have lots of humans involved in building the prompts for the AI that is scripting it, but we don’t have any human involved in creating the audio description on a file-by-file basis. That is all AI supported. And then as I mentioned earlier, we do have a human review process that you can upgrade anything to. So we have both of those options available, but our AI audio description tool does use AI for both the scripting and the voice.
JESSIE ZIONTS: And I did just want to throw in for either the AI audio description or the human-edited audio description, we do have an editing interface for you all, if you wanted to make any adjustments to the scripts that have been created.
LILY BOND: Yeah, great callout, Jessie. Our script editor is a great tool to use, if you don’t want to use 3Play’s human script writers. Great. Another question here is, how is the accuracy if it’s a more complex topic? I think a few questions here are related to accuracy and quality for different niche subjects.
I’m not entirely sure whether this is referring to captioning or audio description, but I think the answer is the same, hopefully, which is that we have worked with higher education for over 15 years. This is where we got our start. We are constantly learning, as a business from all of the content that we produce captions and description for and using that in our AI solutions. So we are able to craft prompts for the description based on the type of content that we’re receiving.
Similarly, for captioning, we are able to use everything that we’ve learned in our history with different types of content to improve our ASR models and have the human corrections, obviously, flowing back into our models moving forward. We certainly would love to test on different types of content. If you have a niche subject that you’re concerned about, we would love to see an example and see how it performs on it. We are eager to test and learn and work with universities to continue optimizing this product. But it’s a great question, and one that we would like to explore together.
What is the max accuracy standard available for closed captions? So our human-produced closed captioning is we guarantee over 99% accuracy. Our measured accuracy rate is actually 99.6%, and that is the leading accuracy rate in the industry.
For ASR or auto captions, it really depends on the content. We produce a study every year, The State of ASR, where we measure the accuracy of the top 10 speech recognition engines to understand how it’s performing for different types of content, which engine is the best for the captioning process specifically. And then we report out on those results. We are in the process of doing that research right now for the 2025 report, and we’ll be sharing that out shortly.
But typically, best case with really good quality audio, not a super niche topic, no accent, single speaker, you could see 95% to 97% accuracy on ASR. For a more realistic view, you’re looking more around like 90% to 92%. And then, as I said, there will be files where it’s 60% accurate, where it’s 70% accurate, because it really depends on the audio characteristics of an individual video.
A great question– have you conducted any research about students’ reactions or learning outcomes on AI versus human accessibility tools? We are in the process of conducting a study with blind students. It’s a user study on our AI-Enabled Audio Description tool, really important to us to get feedback from the end users who require the accommodation. And we will be integrating that feedback into our product and also reporting out on the results once we have them.
Let’s see. Are real-time transcriptions created for course access, then used for AI training? No, we do not use real-time transcripts for AI training. For the Live Summarization, we are creating the transcripts with human voice writers, and then we are using AI not to integrate that into training, but to adapt that transcript into a derivative, which is a summary.
A question here– to meet federal compliance, do all videos, including supplemental and optional content, require audio descriptions, or is the WCAG standard limited to core required instructional materials? That’s a great question. As far as I know, there is no carveout for optional content versus required content. I will follow up on that, but I believe it’s pretty cut and dry that if it’s available, it has to be accessible.
I know there was– initially, universities expected a carveout for course content in the Title II updates, and that did not make the final regulation. And I believe the regulation has very, very few exceptions, and this does not ring a bell as one of them. But I will confirm.
I think that’s all we have time for. So many great questions, really appreciate your engagement. We encourage you, if you are interested in learning more, to schedule a meeting. And always feel free to reach out to me or Jessie with additional questions. Thanks again to everyone for joining. Thank you to Jessie for sharing all of the great feedback from customers and helping present our new products. And I hope everyone has a great rest of your day.
JESSIE ZIONTS: Thanks, everyone. Bye.