1 of 35

2 of 35

��Session Title – �Unleashing the Power of Generative AI for Making Accessible Videos��Speaker Name –�Vijayshree (VJ) Vethantham�SVP, Growth & Strategy, Continual Engine

The topic of today’s session is "Unleashing the Power of Generative AI for Making Accessible Videos", which is an exploration of the importance of video accessibility in today's digital landscape. As technology continues to advance, so does our responsibility to ensure that everyone can fully engage with multimedia content.

I’d like to share an overview about our Invicta™, which a proprietary generative AI solution for automating image alt text. . We leverage Invicta’s caspabilities along with automation and collaborative approaches to accelerate the video accessibility process. This includes closed captions and transcripts but we will be specifically talking about the audio description process which tends to be the most time consuming. We will also highlight the use of AI to automatically generate alt text for complex images and diagrams within videos, making content more accessible for educational content.

3 of 35

Speaker Disclosure

Vijayshree (VJ) Vethantham has relevant Financial relationships to disclose.

(Salary From Employment)

ATIA 2024 Conference | 3

4 of 35

Learning Objectives

Understand AI-based transcription role in inclusive education for learners with disabilities or language barriers.
Discuss Continual Engine's caption management framework, leveraging video processing and NLP for efficient accessibility.
Identify AI's potential in transforming audio descriptions and closed captions, improving efficiency and quality in video accessibility.

ATIA 2024 Conference | 4

5 of 35

Key Takeaways

Understand the role of generative AI for audio descriptions and captions.
Improve accuracy with well-designed AI for video accessibility.
Review the impact of AI in the space of digital accessibility.

ATIA 2024 Conference | 5

6 of 35

Video

Accessibility

ATIA 2024 Conference | 6

7 of 35

How Videos Are Widely Used

In 2022, videos formed 65.93% of the Internet traffic. (IT Web)
YouTube has 2 billion monthly users, and YouTube Shorts hit 50 billion daily views in Feb 2023. (Statista)
Video increased web traffic for 87% of video marketers and sales for 80%. (Hub S pot)
Facebook has 4B video views daily, with 500M viewers watching 100M hours of content. (99firms)
Udemy Business and Skillshare witnesses substantial growth with video courses.
Netflix, Disney+, and Amazon Prime Video were the three most downloaded in 2022. (Statista)

ATIA 2024 Conference | 7

We have increasingly moved into a digital space for all learning, entertainment and communication. Videos are one of the fastest growing formats across the globe. Powerful and easy to create, they still remain largely inaccessible for people using assistive devices.
Let’s look at some data:

In the first half of 2022, video accounted for a hefty 65.93% of total volume over the internet. (IT Web)
YouTube, with more than 2 billion monthly users, is a leading video-sharing platform. Launched globally in June 2021, YouTube Shorts surpassed 50 billion daily views in February 2023. (Statista)
87% of video marketers say that video has increased traffic to their website, and 80% of video marketers claim that video has directly increased sales. (Hubspot)
More than 4 billion video views take place on Facebook every day. 500 million viewers watch 100 million hours of video content daily. (99firms)
Educational platforms like Udemy Business and Skillshare witnessed substantial growth by offering video-based courses.
As of 2022, Amazon’s on-demand service, Amazon Prime Video, was the third most downloaded video streaming mobile app, after Netflix and Disney+ (Statista)

8 of 35

How Video Accessibility Serves �Multiple Purposes

A learning aid
Note-taking tool
Noisy locations
Quiet viewing
Comprehension support
Option for non-native speakers
Contextual understanding
Legal compliance

ATIA 2024 Conference | 8

While the volume of videos continues to increase, there however remains this misconception that video accessibility is optional or not necessary given that not everyone needs accommodation. However, there is a number of reasons for incorporating universal design principles – accommodation being just one of them. Let’s look at how videos are used:

Learning aid: Video accessibility with captions and transcripts helps all learners better understand and remember information, accommodating different learning styles.
Note-taking tool: Transcripts are valuable for referencing and note-taking, making it easier to review and extract key information from videos.
Noisy locations: Closed captions ensure video understanding in noisy environments, where audio quality may be compromised.
Quiet viewing: Closed captions enable silent content consumption in noise-sensitive settings, like libraries or shared workspaces.
Comprehension support: Closed captions assist those with hearing impairments and help individuals struggling with accents, background noise, or language barriers.
Option for non-native speakers: Closed captions make video content more accessible for language learners.
Contextual understanding: Audio descriptions provide essential context for visual elements, enhancing the viewing experience for all.
Legal compliance: Video accessibility is required by law for public-facing content, avoiding legal issues and improving the user experience for everyone.

9 of 35

Growing Challenges With �Video Accessibility

High manual efforts
Time-consuming task
Expensive
Automation impacts accuracy
Backlog of inaccessible educational videos
Relying on subject matter experts
Complex student accessibility needs

ATIA 2024 Conference | 9

10 of 35

Role of Generative AI

Proprietary Generative Artificial Intelligence (AI) is revolutionizing education by harnessing AI's power while safeguarding content integrity and data privacy in a closed environment.

Invicta™, our award-winning generative AI-solution has two user cases-

1. Image alt text

2. Transcription automation

ATIA 2024 Conference | 10

11 of 35

Invicta™ for Video Accessibility �(1 of 2)

Invicta™ leverages proprietary AI technology to create accurate and accessible closed captions and transcriptions for diverse educational content. It's a gamechanger due to its multilingual capabilities, facilitating content processing in various languages.

ATIA 2024 Conference | 11

12 of 35

Invicta™ for Video Accessibility �(2 of 2)

This is On-Screen

Image Content

This is the automated Alt Text created by Invicta™

ATIA 2024 Conference | 12

13 of 35

AI-Assisted Captioning & Transcription Framework (1 of 2)

Continual Engine's Invicta™ uses AI and natural language processing (NLP) to automate and enhance caption and transcript creation, offering businesses and educational institutions a valuable tool for efficient, inclusive communication.

ATIA 2024 Conference | 13

14 of 35

AI-Assisted Captioning & Transcription Framework (2 of 2)

ATIA 2024 Conference | 14

15 of 35

AI-Assisted Video & Audio Description Framework (1 of 2)

Continual Engine's AI-Assisted Video & Audio Description Framework, powered by Invicta™ APIs, advances multimedia accessibility through AI automation, saving time and resources while enhancing accuracy for a more inclusive multimedia experience.

ATIA 2024 Conference | 15

16 of 35

AI-Assisted Video & Audio Description Framework (2 of 2)

ATIA 2024 Conference | 16

17 of 35

Our Approach

Continual Engine receives videos
Invicta™ captions and transcribes.
Invicta™ processes images for alt-text.
SMEs review and make corrections.
Invicta™ adds audio descriptions to videos.
SMEs improve accuracy; which provides algorithms with input/training.

ATIA 2024 Conference | 17

Our unique approach begins with the videos received by Continual Engine, which are initially processed through our AI tool, Invicta, to generate captions and transcriptions. Invicta goes beyond captioning by processing images to produce alt-text descriptions for visual elements. The output from Invicta is then meticulously reviewed by our Subject Matter Experts, who make necessary corrections if any discrepancies are identified. After finalizing the descriptions for visual components, Invicta is employed to generate audio descriptions. Finally, these audio descriptions, along with captioning and transcription files, are seamlessly integrated into the original video content with the assistance of Invicta.

Our approach mitigates the typical challenges linked to full automation in video processing. While full automation can introduce errors due to its limitations in understanding every spoken word or interpreting video content accurately, our reliance on Subject Matter Experts ensures precision and quality. These experts play a critical role in enhancing the accuracy of captioning, interpreting visual elements, and identifying the optimal timing for inserting audio descriptions, areas where AI alone may fall short.

18 of 35

Types Of Videos We Primarily �Work On Using AI

Continual Engine’s Invicta™ primarily works on Science, Technology, Engineering, and Mathematics (STEM) videos primarily from –

Mathematics
Chemistry
Accounting
Management
Finance

ATIA 2024 Conference | 18

19 of 35

Current Benefits And Challenges �(1 of 2)

Current Benefits:

Time Savings: AI enables efficient automation, saving significant amounts of time in various tasks.
Increased Revenue: By streamlining processes, AI contributes to enhanced productivity, leading to potential revenue growth.
Production Boost: Automated tasks and optimized workflows facilitated by AI can result in increased production output.

ATIA 2024 Conference | 19

20 of 35

Current Benefits And Challenges �(2 of 2)

Current Challenges:

Accuracy Limitations with Complex Data: AI may struggle with complex graphs and images when specific context is required.
Human Intervention Needed: Achieving perfection in certain tasks, such as embedding descriptions and captions in videos, may require human review, particularly for precise time stamps.

ATIA 2024 Conference | 20

21 of 35

Limitations Of AI

1. Limited Application in Life-Science and Engineering Images:

AI struggles with the analysis of a significant portion of Life-Science and Engineering images.
Specialized expertise from Subject Matter Experts (SMEs) is often required to provide accurate descriptions for these images.

2. Inefficiency with Historical Pictures and Paintings:

AI faces challenges in efficiently processing and interpreting historical pictures and paintings.

ATIA 2024 Conference | 21

22 of 35

Use Case - Client Examples (1 of 2)

Client: Two of the Big 5 Publishers

Challenge: Work on large volumes of videos, speeding up the process while retaining high-quality output

Approach: A team of 80+ Subject Matter Experts and quality analysts who, along with automation, process large of videos while maintaining high quality.

ATIA 2024 Conference | 22

23 of 35

Use Case - Client Examples (2 of 2)

Invicta™ processed around 750 videos, with a total duration of 3,000 minutes, within the span of one week.

ATIA 2024 Conference | 23

24 of 35

Key Benefits: AI-Assisted �Video & Audio Description

Faster time to market
Lower costs
Easier to scale
More accurate audio descriptions
AI technology + human experts = better quality control

ATIA 2024 Conference | 24

25 of 35

FAQs

1. How do we handle YouTube videos and how do we handle public videos where we may not have source files?

- In cases where a video contains audio, additional context or source files are often unnecessary. Clear visuals also enhance our ability to process videos without audio components effectively. However, if a video uploaded on a public platform requires permission for reuse and accessibility, we prioritize obtaining the necessary permissions before making it accessible, ensuring compliance with legal and ethical considerations.

How can this technology be made available to educators and organizations?

- Currently, this technology is not available for off-the-shelf use by educators or organizations. As the accuracy of AI improves, CE will begin making the video accessibility technologies available to clients and partners.

ATIA 2024 Conference | 25

26 of 35

FAQs

3. How much manual intervention is needed?

- Automation and AI significantly expedites the overall process in video accessibility but human experts are consistently involved at the end of each stage of the workflow, to ensure accuracy and bring in contextual information.

4. What are the types of videos that are well-suited for AI?

- Currently, the accuracy of AI is highest in STEM and quantitative disciplines. However, the technology is fast advancing to automate diverse images in other disciplines like management, finance, biology, anatomy, and more.

ATIA 2024 Conference | 26

27 of 35

About

Continual

Engine

28 of 35

Introduction

Continual Engine (www.continualengine.com) is an award winning artificial intelligence (AI) technology company dedicated to building smarter, affordable, and scalable solutions for learning and accessibility.

ATIA 2024 Conference | 28

29 of 35

Awards Won

ATIA 2024 Conference | 29

30 of 35

Continual Engine: �Solutions & Services

Accessibility Services

Document Remediation (PREP)

Image Alt Text (Invicta™)

Automated PowerPoint Accessibility

Automated Video Closed Captions & Audio Descriptions

Website Accessibility

Learning Process Automation

Digitization

Tagging, Indexing, and Search

Learning Aids/Assessments

ATIA 2024 Conference | 30

31 of 35

Continual Engine’s�Journey Into Accessibility & Learning

Decades of tech and content expertise.
Transforming accessibility with tech and experts.
Award-winning solutions with AI.
Analyzing complex content for automation.
A team of tech experts, engineers, and more.

ATIA 2024 Conference | 31

32 of 35

Continual Engine: Compliance Testing �& Quality Checks

Customizable solutions for client guidelines.
High-quality output in various formats.
Up to 60% time and 50% cost savings with scalability.
In-house experts verify accessible versions.
Exceeds standards like WCAG 2.1/2.2 AA, PDFU/A, and ADA.

ATIA 2024 Conference | 32

33 of 35

Partnering with Continual Engine

We offer subscription packages and customized services for accessibility including image descriptions, document remediation, video closed captions and extended audio description services, and more.

ATIA 2024 Conference | 33

34 of 35

Questions

35 of 35

Thank You

Questions or Feedback?

Vijayshree (VJ) Vethantham

SVP, Growth & Strategy

Chicago, IL

vijayshree.vethantham@continualengine.com