You can remove filler words in Descript in about two minutes, and most of that time goes to reviewing what it found rather than clicking. Descript transcribes your recording, underlines every “um” and “uh” in the transcript, and cuts the matching audio and video when you tell it to. Nothing in the process involves scrubbing a waveform by hand.
The tool is fast, but the quality of the result depends on three choices that most walkthroughs skip. The first is what you do with each filler: delete it, ignore it, or remove it from the transcript only. The second is whether you leave the Avoid harsh cuts setting on. The third is whether you correct the transcript before you run the removal. This guide covers each choice, what the feature costs in AI credits, and the situations where leaving fillers in produces a better video.
How Descript Finds Filler Words
Descript edits video through its transcript. When you import a recording, it transcribes the whole file and links every word to a position in the audio and video. Because each word has a timestamp, deleting a word from the text deletes the matching slice of the recording.
Filler word removal builds on that. Descript scans the transcript for common fillers such as “um,” “uh,” and “you know,” and marks them with a light blue underline in the script editor. When you run the removal tool, it acts on those marked words. The detection happens at the text level, so the accuracy of the removal depends on the accuracy of the transcript. If the transcription model hears a filler that you did not say, or misses one that you did, the cut follows the transcript rather than your actual speech.
That is the reason the correction step below matters. Users in Descript’s own feedback forum describe the removal as hit or miss on messy audio, and several say they now check the results manually instead of bulk-removing everything. The tool saves time, but it does not replace a listen-through.
Step by Step: Removing Filler Words in Descript
- Import or record your video. Create a project and drop in the file, or record directly in Descript. Transcription starts automatically. On current plans the imported minutes count against your monthly media minutes.
- Read the transcript once before you touch anything. Fix misheard words, especially short ones. A wrong word in the transcript becomes a wrong cut later.
- Open the AI Tools panel. It sits in the right sidebar. Choose Remove filler words.
- Review the list. A sidebar opens with every filler word Descript detected. You can pick which ones to act on, so you can remove “um” and “uh” and leave “you know” in if it fits how you talk.
- Choose an action for each filler. A dropdown offers Delete, Ignore, or Remove from transcript. The table in the next section explains the difference.
- Decide on Avoid harsh cuts. Leave it on unless you have a reason to turn it off. It is covered in its own section below.
- Apply the removal. Descript makes the cuts across your composition.
- Play the result. Listen to the full video at normal speed with headphones on. Mark any spot that sounds clipped and fix it by hand.
By default the removal applies to the entire composition. If you only want to clean one section, such as the first two minutes of a longer recording, put that section in its own composition and run Remove filler words there.
Delete, Ignore, or Remove From Transcript: What Each Option Does
The action dropdown is where most mistakes happen, because the names do not describe the outcome as clearly as they could. Descript’s help center defines them like this:
| Option | What happens to the text | What happens to the audio | Use it when |
|---|---|---|---|
| Delete | Removed from the script | Removed | You want the filler gone completely, in both the transcript and the video |
| Ignore | Stays in the script with a strikethrough | Removed | You want the audio cut but a visible record of what was removed |
| Remove from transcript | Removed from the script | Unchanged | You want clean captions but natural-sounding audio |
For most videos, Delete is the right default. The third option is more useful than it looks. A Descript video editor describes it in one of the company’s own tutorials: cutting every filler from the audio can leave speech sounding choppy, while leaving fillers in the caption text clutters the captions. Removing the words from the transcript only gives you readable captions over audio that still sounds like a person talking.
What Avoid Harsh Cuts Does
Avoid harsh cuts is a toggle in the same sidebar. When it is on, Descript analyzes the audio around each filler and skips any that cannot be removed without clipping a neighboring word or leaving an awkward edit. In other words, the tool deliberately leaves some fillers in.
That has a practical consequence. If you run the removal with the setting on and then notice a few “ums” still in the video, the tool did not miss them. It judged that cutting them would sound worse than keeping them. You can remove those by hand if you disagree, usually with a small trim or a short crossfade in the editor.
Turn it off only if you are cleaning audio that will be heavily processed afterward, or if you plan to review every cut manually anyway. Feedback posts from Descript users show the main complaint about bulk removal is clipping, and this setting exists to address it.
Which Filler Words Descript Detects
The help center says Descript detects common fillers such as “um,” “uh,” and “you know.” The current version of the help page lists English, German, French, Portuguese, and Italian transcripts as supported. An older version of the same page listed English only, so if you edit in another language, test the feature on a short clip before you rely on it for a full project.
Descript’s help page also notes that people on legacy plans have a different feature set. Creator (Legacy) plans include the complete list of filler words plus repeated words, while other legacy tiers are more limited. If you subscribed before the 2025 pricing change, check your own plan’s behavior instead of assuming it matches the current documentation.
What It Costs: AI Credits and Media Minutes
Descript meters two things. Transcription and editing draw on media minutes, which count the minutes of audio or video you import or record. AI features draw on a separate pool of AI credits. According to the help center, Remove filler words uses AI credits on current plans. The help page does not list a per-run credit cost, so the practical way to learn what a run costs is to check the usage tracker in your account before and after one removal.
| Plan | Billed monthly | Billed annually (per month) | Media per month | AI credits |
|---|---|---|---|---|
| Free | $0 | $0 | 60 minutes | 100, one time |
| Hobbyist | $24 | $16 | 10 hours | 400 per month |
| Creator | $35 | $24 | 30 hours | 800 per month |
| Business | $65 | $50 | 40 hours | 1,500 per month |
Prices are per person and were verified on Descript’s pricing page in August 2026. They change, so confirm them before you subscribe.
The free plan is enough to test the feature. It includes 60 media minutes per month and a one-time grant of 100 AI credits, but exports are 720p with a Descript watermark. The first paid tier removes the watermark and exports at 1080p. 4K export starts on Creator. The 100 free credits do not renew, so a new free account can run the tool a limited number of times before it needs a paid plan.
One more detail from third-party pricing breakdowns: annual billing is paid for the year up front, and one comparison reports that top-ups for extra media minutes and credits are only available on Creator and Business. If you expect to run out mid-month, that affects which tier you pick.
Which Filler Words to Remove and Which to Keep
Removing every filler is not always the goal. Descript’s own editors say filler words carry pacing information, and that removing all of them from conversational content can make it sound stiff. A talking-head tutorial and a two-person podcast need different treatment.
| Content type | Recommended approach | Reason |
|---|---|---|
| Short-form talking-head video | Remove almost all, keep Avoid harsh cuts on | Every second counts and viewers decide within the first few seconds whether to keep watching |
| Product demo or tutorial | Remove “um” and “uh,” keep short natural pauses | Viewers need time to follow what is on screen |
| Interview or podcast | Remove the obvious ones, leave conversational ones | Over-editing makes the exchange sound scripted |
| Course lesson | Remove most, review each cut | Polish matters, but a flat delivery loses students |
| Testimonial or story | Remove sparingly | Hesitation can read as honesty |
A useful test is to listen for the type of filler. A filler that fills dead air while you search for the next word is safe to cut. A filler that signals a real pause for thought, right before a point lands, often helps the delivery. The tool cannot tell these apart, which is why reviewing the list before applying it is worth the extra minute.
Pacing also affects how a short video performs, because retention is one of the signals a platform uses when it decides how widely to show a video.
How Much Shorter Will the Video Get?
Less than most people expect. One creator-education tutorial estimates that a 15-minute recording with around 40 filler words loses 30 to 60 seconds after removal. That is an estimate from a single source rather than a Descript benchmark, but it matches the arithmetic. A filler word lasts a fraction of a second, and removing it also removes some surrounding silence.
Scaled to a 10-minute talking-head video with around 30 fillers, a rough expectation is 20 to 40 seconds saved. That tightens the delivery, but it does not turn a rambling 90-second clip into a 60-second one. If your recording is too long, the fix is cutting content, not fillers.
Users also mention a separate Descript tool that shortens long gaps between words. People who want a faster pace often run it after filler removal, and several in the feedback forum say they review both tools’ cuts by hand rather than applying them in bulk.
A Walkthrough: A 10-Minute Product Video
The numbers in this example are illustrative, not measured, but the workflow is the one described above.
- You record a 10-minute talking-head review of a kitchen gadget. Descript transcribes it and underlines 31 filler words.
- You read the transcript and fix four misheard words, including one place where “a” was transcribed as “uh.” Left uncorrected, that one error would have cut a real word.
- You open AI Tools, choose Remove filler words, and uncheck “you know,” because you say it once in a sentence where it sounds natural.
- You set the action to Delete and leave Avoid harsh cuts on. Descript removes most of the fillers and skips a handful it judged too risky to cut.
- You play the result with headphones. Three spots sound abrupt, so you trim each by a few frames.
- You export a 1080p file, roughly 25 seconds shorter than the original, with the pacing noticeably tighter.
The whole process takes less time than the first step of manual editing would have, which is finding the first filler on the timeline.
Common Problems and How to Fix Them
| Problem | Likely cause | Fix |
|---|---|---|
| Audio sounds choppy after removal | Fillers cut without enough room around neighboring words | Undo, rerun with Avoid harsh cuts on, then trim the leftover spots by hand |
| A real word disappeared | The transcript misidentified it as a filler | Undo, correct the transcript, and rerun |
| Fillers still in the video | Avoid harsh cuts skipped them on purpose | Remove them manually if the cut sounds acceptable |
| Captions look cluttered | Fillers were left in the transcript | Use Remove from transcript to clean the text without touching the audio |
| Visible jump in the video at a cut | Talking-head footage with no cutaway | Add a zoom, a b-roll cutaway, or a short transition over the edit |
| Tool fails or asks for credits | AI credit balance is empty | Wait for the monthly reset, buy a top-up if your plan allows it, or upgrade |
Cleaner Captions Without Choppy Audio
The Remove from transcript option solves a problem that many short-form creators run into. Auto-generated captions that include every “um” look messy on screen, but audio with every filler removed can sound unnatural. Run the filler removal tool, pick the words you want gone, and change the action from Delete to Remove from transcript. Your captions come out clean and your audio keeps its natural rhythm.
This works best for videos where the speaker’s voice is the main draw, such as commentary or reviews. For videos with fast visual editing, deleting both the audio and text usually fits better.
Reduce Fillers Before You Record
The easiest filler to remove is the one that never got recorded. Descript’s own team suggests that knowing you can edit fillers out lets you stop worrying about them and just speak a bit more slowly. Beyond that, scripting your key lines in advance cuts hesitation, because most fillers appear when you are deciding what to say next. If you write scripts with an AI tool, a quick read-aloud pass before filming catches the awkward phrasing that causes stumbles.
Batch Editing Several Videos in One Session
If you film several videos at once, run the filler removal as a separate editing pass rather than cleaning each video right after you record it. Record everything, import all the files, correct each transcript, then run Remove filler words on each project in sequence. Doing the same task repeatedly keeps your listening standard consistent, so you catch the same kinds of problems across every video.
Planning that batch ahead of time also helps. Deciding what you will film for the month makes it easier to keep each recording short and focused, which means fewer filler words to remove in the first place.
Clean the Recording, Then Cut the Clips (Cheat Code)
A cleaned recording is more useful than a single published video. A 10-minute review with filler words removed can supply a full-length upload, a few shorter cuts, and the raw material for different angles on the same product. That kind of repurposing is how a small number of recordings can fill a posting schedule.
If your source material is a longer recording, such as a podcast episode or a livestream, the next question is how to find the best short moments inside it. Automatic clipping tools can do that search for you.
Filler removal handles one part of a long recording workflow: making the audio tight. The other part is deciding which 30 to 60 seconds of a 40-minute recording deserve to become a short video. Scrubbing through the timeline to find them is slow, and it is the step that most creators skip, which is why long recordings sit unused.
OpusClip takes a long video and finds candidate clips automatically, ranks them by a predicted-performance score, and exports them captioned and reformatted for vertical video. A workable order is to clean the recording in Descript first, so the audio is tight, then feed the finished file into OpusClip and review the top-ranked clips before you post anything. The free plan includes 60 credits, which is enough to run one full test episode and judge the output for yourself.
Is Descript the Right Tool for This?
Worth using if:
- You record talking-head videos, tutorials, or podcasts and spend real time cutting ums by hand
- You are comfortable listening through the result once before exporting
- You publish often enough that the monthly cost is offset by the editing time saved
Probably not worth it if:
- Your videos are short, tightly scripted, and already free of fillers
- You need frame-accurate control over every cut and prefer a traditional timeline
- You only edit a few minutes of footage each month, in which case the free plan may cover you
Descript vs. Manual Editing
| Descript | Manual timeline editing | |
|---|---|---|
| Time to find fillers | Seconds, since they are underlined in the transcript | Requires listening to the full recording |
| Precision of each cut | Good, but occasionally clips neighboring words | As precise as you make it |
| Cost | Free tier available, paid plans from $16 per month on annual billing | Free if you already own an editor |
| Learning curve | Low, since it works like editing a text document | Higher, especially for audio crossfades |
| Best for | Regular publishing and long recordings | One-off projects or audio that needs fine control |
Some experienced editors prefer to cut fillers by hand with a dedicated audio tool because they want to control every edit. A Descript forum thread from one such user describes the built-in removal as hit or miss next to manual editing in a professional audio restoration program. Both approaches work. The question is whether your time is better spent on the cut or on the content.
Frequently Asked Questions
Does Descript remove filler words automatically?
It detects them automatically and underlines them in the transcript. Removal happens when you run Remove filler words from the AI Tools panel, and you choose which fillers and what action to apply.
Is filler word removal free in Descript?
The free plan includes a one-time grant of 100 AI credits, and the help center says the feature uses AI credits on current plans. You can test it for free, but ongoing use needs a paid plan or additional credits.
Will removing filler words change the length of my video?
Yes, but only slightly. One estimate puts it at 30 to 60 seconds saved on a 15-minute recording with about 40 fillers. Gaps close where the fillers were, so the video gets a little shorter and tighter.
Can I remove filler words from only part of a video?
Yes. The removal applies to the entire composition by default. To limit it, place the section you want to clean in its own composition and run the tool there.
Why are some filler words still in my video after I ran the tool?
If Avoid harsh cuts is on, Descript skips any filler it cannot remove without clipping a nearby word or creating an awkward edit. You can remove the remaining ones by hand.
Can I clean up captions without changing the audio?
Yes. Set the action to Remove from transcript. The word disappears from the script and captions, but the audio is unchanged.
Does it work in languages other than English?
The current help page lists English, German, French, Portuguese, and Italian. An older version listed English only, so test a short clip before using it on a long project in another language.
What if the tool removes a word I wanted to keep?
Undo the edit, correct the transcript if the word was misidentified, and rerun the removal. Reviewing the detected list before you apply it prevents most of these cases.
Pre-Export Checklist
- Transcript corrected before running the removal
- Filler list reviewed, with natural-sounding fillers unchecked
- Action set to Delete, or Remove from transcript if you only want clean captions
- Avoid harsh cuts left on unless you have a reason to turn it off
- Full video played at normal speed with headphones
- Abrupt cuts trimmed or covered with a zoom, cutaway, or transition
- AI credit balance checked before the next batch
The Bottom Line
Descript removes filler words quickly and with little effort, and the free plan is enough to test it on a real video. The result is as good as the transcript behind it and the settings you choose. Correct the transcript first, review the list of detected fillers, use Remove from transcript when you only need clean captions, and listen to the final cut before you export. Treat the tool as a fast first pass that handles most of the work, and finish the last few cuts by ear.
Team82 is the Flux82 editorial team focused on short-form affiliate education, TikTok Shop creator workflows, platform behavior, content systems, and conversion mechanics. Flux82 publishes practical guides for creators who want clearer execution frameworks, better posting systems, and more structured ways to understand how short-form affiliate content works. Follow Flux82 on X at https://x.com/Flux82Lab.