Reproduce “DescribePro: Collaborative Audio Description with Human-AI Interaction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Maryam Cheema, Sina Elahimanesh, Samuel Martin, Pooyan Fazli, Hasti Seifi
DescribePro introduces a web-based platform for collaboratively authoring audio descriptions (ADs) through human-AI co-creation and human-human interaction. The system enables describers to refine AI-generated drafts using natural language prompts and create diverse AD variations, achieving an average System Usability Scale score of 72.6 and demonstrating efficiency in generating accessible video content.
Key insights
- Combining AI efficiency with human expertise through a human-in-the-loop approach successfully balances quality and scalability in AD authoring.
- Collaborative features like forking promote the creation of diverse narrative styles and foster community-driven improvement, addressing varied user preferences.
- Professional and novice describers have distinct needs: professionals leverage AI for targeted linguistic refinement, while novices benefit from AI as a confidence-building starting point and learning aid.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.