Download

Download Full Text (368 KB)

Description

As the April 24, 2026 ADA Title II deadline approaches, higher-education instructors are facing the task of improving captions on existing course videos. This poster presents a practical workflow to improve Canvas Studio captions in which AI performs most of the cross-checking between two independent transcripts: the original Studio captions and an AI-generated transcript from the extracted audio. The AI compares the transcripts and uses differences as a quality-check signal: segments with strong agreement remain unchanged, while mismatches, “inaudible/uncertain” markers, and glossary terms (names, acronyms, technical vocabulary) are automatically flagged for targeted human review. After review, the workflow outputs a revised caption file, optionally removing filler words, while preserving timing and updating caption text where warranted. The goal is a scalable, instructor-friendly approach that concentrates human effort where it adds the most value.

Department of Primary Author

Technology & Workforce Learning

Affiliation of Primary Author

Faculty

Publication Date

2026

Correct my Captions: An AI-Human Collaboration Workflow

Share

COinS