Improve transcript-based editing accuracy and reliability

The challenge
Mozaik already supports transcript-based video editing: users can select specific words or sections in the transcript of their video footage and delete them, and the editor automatically removes the corresponding segment from the video timeline.

However, the current transcript editor output is not accurate or reliable enough in some editing workflows:

• Cuts are often too aggressive: When users delete specific words or sections from a statement, the resulting edit is frequently too tightly attached to the selected words. Users usually still need to fine-tune the cut manually before it is usable.

• Longer videos can become inaccurate or async: With larger video assets ranging from multiple minutes to several hours, the transcript editor can become slightly out of sync with the actual video timeline, making transcript-based edits less reliable.

Our approach
Improve the precision and reliability of transcript-based editing so that deleting words or sections from the transcript creates more accurate timeline cuts by default, with better handling of longer video assets and fewer cases where users need to manually fine-tune the resulting edit.

Expected impact

• Faster text-based video editing with fewer manual corrections

• More reliable transcript-to-timeline synchronization, especially for longer assets

• Cleaner default cuts when removing individual words, phrases, or sections

• Higher confidence in using transcript editing as a primary editing workflow