Lesson 7 of 13 · Module 3: AI-Assisted Post-Production
Text-Based Editing
Cut a rough assembly of a spoken clip by editing its transcript text, removing filler and dead air, rather than scrubbing a timeline by eye.
Pick: “Descript AI Tutorial | Text-Based Video Editing Made Effortless
Direct Descript match, short and recent, walks through editing a video by deleting transcript text with the automatic ripple. Good on-screen demonstration of the exact workflow.
Objective
Why This Matters
Text-based editing is the actual shift underneath the tool names: you edit by reading and deleting words, not by hunting through a waveform for the right frame. It’s dramatically faster once it clicks, and it’s the direction the entire category has moved regardless of which specific product wins next. Reading a transcript and deleting a line is a task you already do daily with written text. This lesson just points that exact skill at audio.
The Technique
Import or transcribe the clip. Read the transcript like you’d read a document, not like you’d scrub a timeline, and delete: filler words (um, uh, false starts), redundant restatement of the same point, and dead air between sentences that adds nothing.
Cut precisely at word boundaries, never mid-syllable. Most text-based editors will ripple the audio automatically when you delete a word or phrase, keeping sync without you touching a waveform at all, that automatic ripple is the entire point of working this way. Use the waveform view only as a final check, to confirm a cut point sounds clean, never as the tool you use to make the cut in the first place.
Watch the balance between tightening and gutting. The goal is to remove what adds nothing, not to remove everything that isn’t essential. A transcript cut so aggressively that the remaining sentences no longer connect logically is a failure even if the duration target is hit.
Watch For This
Good
- Cuts land cleanly at word boundaries with no audible click or stutter.
- The remaining content reads and sounds like a natural, connected statement.
- Meaningfully tighter than the source, with the reduction coming from filler and dead air, not from gutting content.
Classic Failure
- A cut lands mid-word or mid-syllable because the transcript boundary you clicked didn’t line up with the actual audio waveform boundary, producing an audible click.
- So much was removed that the remaining sentence doesn’t logically connect to what follows it.
- The waveform was used to make the cut instead of the transcript, defeating the entire purpose of the workflow.
Your Drill
Take one raw spoken clip of at least 60 seconds. Cut it by editing the transcript: delete filler, false starts, and dead air. Submit the source clip’s length, the cut clip’s length, the percentage reduction, and the cut clip itself.
Done? Paste what you made into the AI coach below for notes against this lesson's pass checklist.
Pass Checklist
Lesson complete
Criterion met: The resulting cut is at least 20% shorter in duration than the source, contains no audible mid-word cuts or clicks at any edit point, and every cut can be pointed to as a specific deleted line in the transcript, confirmed on one full-volume playback.
Next: Lesson 8: AI Audio CleanupHow solid did that feel?
Noted.
Coach Note
The first time you delete a line and the audio ripples closed automatically with no click, that’s the moment this workflow clicks for good. If you find yourself reaching for the waveform to make a cut instead of just checking one, you’re still editing the old way. Force yourself back to the transcript.
AI Coach
Conversations clear when you leave the page.
Ask about this lesson, or paste what your drill produced above and get it checked against the list.
The coach comes online shortly.
Resurfaces In
Lesson 8 (AI Audio Cleanup), Lesson 12 (Pipeline), Lesson 13 (Capstone).