Making short-form videos accessible for the Blind and Low Vision Viewers

B2C

B2C

Accessibility

Accessibility

AI

AI

UX Research

UX Research

Mobile

Mobile

Launched

Launched

Short-form videos on TikTok, Instagram Reels, and YouTube Shorts remain inaccessible to millions of blind and low vision (BLV) users. I joined the Human-Computer Interaction Lab at the University of Texas at Austin to help address this challenge.

Timeline

Jun 2023 - Sep 2023

My Role

User Researcher & Co-Author

Contribution:

UX Data Analysis (100%)

User Interface Design (60%)

Pipeline Evaluation (50%)

Team

2 User Researchers

4 Developers

Impact

Launched (Open Source)

2.7x Comprehension Accuracy

2.2x Browsing Usefulness

CHI 2024 Publication

Reported by 'The Daily Texan'

Timeline

Jun 2023 - Sep 2023

My Role

User Researcher & Co-Author

Contribution:

UX Data Analysis (100%)

User Interface Design (60%)

Pipeline Evaluation (50%)

Team

2 User Researchers

4 Developers

Impact

Launched (Open Source)

2.7x Comprehension Accuracy

2.2x Browsing Usefulness

CHI 2024 Publication

Reported by 'The Daily Texan'

Timeline

Jun 2023 - Sep 2023

My Role

User Researcher & Co-Author

Contribution:

UX Data Analysis (100%)

User Interface Design (60%)

Pipeline Evaluation (50%)

Team

2 User Researchers

4 Developers

Impact

Launched (Open Source)

2.7x Comprehension Accuracy

2.2x Browsing Usefulness

CHI 2024 Publication

Reported by 'The Daily Texan'

Solution: A summarization system that supports BLV viewers in browsing and understanding short-form videos

Solution: A summarization system that supports BLV viewers in browsing and understanding short-form videos

Challenge

Making short-form videos accessible for the Blind and Low Vision (BLV) Viewers

Today, 1+ billion users watch short-form videos on TikTok, Instagram Reels, and YouTube Shorts. However, based on the literature review, they remain inaccessible to millions of blind and low vision (BLV) viewers due to their rapid visual changes, on-screen text, and music or meme-audio overlays.

❌ Problem

  1. On-screen text is hard to read by screen reader.

  1. Rapid Visual Change.

  1. Insufficient description uploaded by the author.

  1. Music Audio is irrelevant to the video content.

Research

Short-form videos on TikTok rated only 3.2/7 in accessibility. I uncovered key insights through 7 in-depth user interviews to understand why.

To uncover the extent, contexts, and causes of inaccessibility in short-form videos, I analyzed data from seven 1.5-hour interviews with BLV participants, each featuring two video-watching tasks. The average rating of 3.21 shows these videos are less than halfway accessible. 

Insight 1 Unlike sighted users, BLV users can’t quickly determine a video's content and decide whether to continue watching it.

❌ Problem

  1. On-screen text often lacks context or clarity.

  1. Incomplete descriptions hinder content understanding.

  1. Hashtags provided by creators may not accurately reflect content.

Insight 2 — BLV users face difficulties comprehending various short-form video formats due to inconsistent or inadequate metadata and descriptions.

  1. Micro-length videos (less than 5s) had uninformative or ambiguous audio that interfered with screen reader audio.

  1. Repurposed audio lacks context due to insufficient descriptions.

Insight 3 — BLV Users relying on VoiceOver encounter significant accessibility barriers due to poorly labeled video control buttons and frequent changes in the interface layout. 

  1. With VoiceOver enabled, video control buttons (e.g., play, pause) are hard to trigger

Design

Providing video descriptions emerged as a straightforward and effective design solution given their compatibility with screen readers

  • I designed a short description of video content so that BLV users can quickly access essential information and make decisions. -> Insight 1

  1. Short Description

10-word description condensed from the long description of the video.

  1. Description Pane Button

Click to enter the description pane, which is the second feature below.

  • I implemented consistent, multi-level descriptions compatible with screen readers to ensure all users can easily understand and engage with the content. -> insight 2

  1. Long Description

A paragraph summarizing the video.

  1. On-Screen Text

Extracted from the video.

  1. Shot-by-shot Description

Include descriptions of each shot in the video, which provide more granular information.

  • I standardized button labels, ensuring compatibility with screen readers, and minimizing unnecessary layout change. I also redesign the feature layouts for quicker access and more intuitive navigation. -> Insight 3

  1. Button Layout Redesign

Highlight video control buttons, and organize the remaining buttons based on user preferences.

Final Solution

An interface that allows users to access video descriptions at four levels of detail based on interest

I developed ShortScribe, which features a video pane and a description pane. 

The video pane (built upon existing short-form video platforms) includes 

  • A short description feature.

  • A redesigned button and information layout tailored to BLV users' preferences.

  • A new button to access the description pane feature. 

The description pane (unique to our interface) provides

  • Hierarchical video summaries at three levels of detail - long description, on-screen text, and shot-by-shot description.

This allows BLV viewers to consume the depth of information based on interest.

This allows BLV viewers to consume the depth of information based on interest.

🤩

🤩

Video Showcase

User Evaluation

Short-form videos became measurably more accessible for BLV users with our system

After pipeline evaluation, I conducted a within-subjects study where participants compared our system, ShortScribe, to a baseline interface for watching short-form videos.


All of ShortScribe's description features—long, short, shot-by-shot, and on-screen text—were rated significantly more useful than the baseline information.

  • For video selection, users rated our system 2.2x more useful than the current one. 

  • For video comprehension, participants rated our system 2.7x more useful than the current one. 

  • Participants praised the accessibility of explicit video controls in our interface.

Design improvements

I made 3 improvements to the system, based on the user testing

Feedback 1

Despite high accuracy in pipeline evaluation, some errors in the descriptions persisted, often unaware by users. 

Solution 1

Add a setting option to display accuracy scores next to descriptions, helping users gauge their reliability.

  1. Accuracy

Display accuracy near description type if the user turns on ‘show accuracy’ in the setting.

Feedback 2

The design inevitably created some redundancy in descriptions, both with the original video audio and across different summaries. However, our usability test revealed some users preferred having more information rather than less.

Solution 2

Add a customization icon at the top of the description panel, allowing users to tailor and save their preferences for all future descriptions.

  1. Prompt Customization Button

  1. Text or Audio Input the Prompt

B. Updated descriptions based on the prompt

C. Save Prompt Settings

Users can save these prompt settings for future use

Feedback 3

Participants wanted to know video-length information to help them make informed choices when selecting videos.

Solution 3

Added a setting option to display video length on the Video Pane.

  1. Video Length

Display video length near short description if the user turns on ‘show video length’ in the setting.

Impact

Overall, users rated our system 2.5x more accessible compared to the baseline system

2.7x

Video Comprehension Accuracy

Video Comprehension Accuracy

Comprehension Accuracy

2.2x

Video Browsing Usefulness

Video Browsing Usefulness

95.7%

Willingness to Use

Willingness to Use

“It would give me a whole new avenue […] I would even pay for it”

— Participant 6

“It would give me a whole new avenue […] I would even pay for it”

— Participant 6

The paper I co-authored was published on CHI 2024 and featured in The Daily Texan newspaper

The paper I co-authored was published on CHI 2024 and featured in The Daily Texan newspaper

CHI 2024 Publication

Daily Texan Report

UT CS Homepage

Create a free website with Framer, the website builder loved by startups, designers and agencies.