Dynamic Authoring of Audio with Linked Scripts
summary
A system called Voice Script links written script and recorded audio into a single master script that merges script text with ASR transcripts (uses IBM speech-to-text) and per-word timestamps so edits to text propagate to audio. It aligns segments with a Needleman-Wunsch based word match and an alternating segmentation heuristic, lets the user accept segment overrides audio across multiple takes, and supports mixing takes. Evaluation includes informal and comparative user studies. Limitations: ASR errors, only asynchronous collaboration, and non-speech audio isn’t represented.
AI-written summary · about
chapters (8)
- 0:00Start
- 1:11Separation makes the tasks harder.
- 2:58Key Idea 2
- 8:33System Overview
- 9:05Challenge
- 12:48Matching segments
- 14:38Evaluation Highlights
- 16:52Summary
related talks
20mNon-Linear Editing of Text-Based Screencasts
07mAutomatic Instructional Video Creation from a Markdown-Formatted Tutorial
14mAligning Basics of Intelligent Contents for Beginners
12mSynthia: Visually Interpreting and Synthesizing Feedback for Writing Revision
39mCarving Text at Its Joints: A New Perspective on Writing and Computers
30mDocumentation with Human Connection
48mAchieving Optimal Accessibility in Storyline and Other eLearning Environments
09mDan Woods interviewed at TOC 2011