AutoDub: Automatic Redubbing for Voiceover Editing
summary
AutoDub automatically locates and replaces misread segments by using subsequence Dynamic Time Warping on a pair of recordings, then performs crossfading to blend the new snippet into the original. Informal tests show an edit of n seconds needs roughly n seconds of context on each side, and a user study finds AutoDub is faster than expert manual editing as tracks grow and lets novices complete redubbing in about the same time as experts, with uniformly higher satisfaction.
AI-written summary · about
chapters (7)
- 0:00Introduction to AutoDub
- 1:07Problem with manual redubbing
- 2:31Automating redubbing with DTW
- 4:40Demo of AutoDub in action
- 6:46User study results
- 7:49Summary and future work
- 8:36Q&A session
related talks
14mRescribe: Authoring and Automatically Editing Audio Descriptions
21mQuickCut: An Interactive Tool for Editing Narrated Video
07mAutomatic Instructional Video Creation from a Markdown-Formatted Tutorial
19mVidCrit: Video-based Asynchronous Video Review
20mNon-Linear Editing of Text-Based Screencasts
24mFail fast, fail cheap, making voice apps the guerrilla way
06mHierarchical Summarization for Longform Spoken Dialog
17mBest AI Tools For Life & Business