How to Remove Filler Words Automatically in Descript

Nothing makes a talking-head video feel more amateur than a steady drip of ums and uhs. Descript can detect and delete them across a whole recording with a couple of clicks, then let you review what it caught. This guide shows the full workflow plus how to avoid over-cutting.

What you need

• A Descript project with a transcribed video or audio file
• A recording where you actually say filler words (most do)
• A few minutes to review the results

Step 1: Open the Filler Words tool

With your transcript open, go to the toolbar and find Edit, then Remove filler words. On newer builds it lives under the Underlord or magic tools panel. Descript scans the transcript for known filler tokens such as um, uh, like, you know, and so.

Step 2: Choose which fillers to remove

Tick only the categories you trust. Um and uh are almost always safe. Be careful with like, so, and you know, because those are real words in many sentences. Removing every so can mangle phrasing, so leave the risky ones unchecked on your first pass.

Step 3: Run the removal

Click Remove. Descript marks every matched filler as deleted across the transcript and trims the matching audio and video. The deletions show up as edits in your document, so the cut is non-destructive to your source media.

Step 4: Listen back and fix awkward cuts

Play through the affected sections. A removed um occasionally leaves a hard splice or clipped breath. If a cut sounds rushed, click into the transcript at that point and either restore a tiny pause or use the Studio Sound effect later to smooth the audio.

Result: a podcast episode that had roughly 90 ums and uhs now plays smoothly after one tool pass and a five minute review. The transcript also reads cleaner for anyone who reads captions.