Blog

50 Video Editing Prompts That Actually Work

The Blob Team5 min readUpdated September 12, 2026

Fifty plain-English editing instructions grouped by job (pacing, captions, audio, b-roll, framing and export) plus the patterns that separate a prompt that lands from one that misses.

When the interface is a conversation, the skill is not operating software. It is knowing what to say.

Below are fifty instructions that reliably produce good results, grouped by what they do. Use them as-is or as templates. After the list there is a short section on *why* certain phrasings work, which is more useful than the list itself.

Pacing and cuts

The foundation of any edit. Start here, always.

  1. 01Remove the filler words and dead air.
  2. 02Cut every pause longer than a second.
  3. 03Cut it tighter, I want high energy.
  4. 04Leave a little breathing room, it feels rushed.
  5. 05Cut the part where I fumble the intro.
  6. 06Remove the section where I go off on a tangent.
  7. 07Trim the beginning so it starts right when I start talking.
  8. 08Keep the pause before the punchline.
  9. 09Cut to the moment I say "here's the trick."
  10. 10Speed up the boring middle section slightly.

Captions and text

  1. 01Caption it TikTok style, word by word, big and bold.
  2. 02Add clean subtitles at the bottom for long-form YouTube.
  3. 03Highlight each word as it's spoken.
  4. 04Make the captions bigger and centre them.
  5. 05Add a subtle dark background behind the captions so they stay readable.
  6. 06Make the highlight colour pink to match my brand.
  7. 07Put a title card at the start that says "How I Edit in 5 Minutes."
  8. 08Add a lower-third with my name and handle.
  9. 09The captions spelled a name wrong, fix it to "Okafor."
  10. 10Move the captions up so they don't overlap the app UI.

Music and audio

  1. 01Add lo-fi music that ducks under my voice.
  2. 02Put upbeat background music under this, quiet enough to talk over.
  3. 03Make the music quieter, it's competing with my voice.
  4. 04Add a calmer track, this one's too intense.
  5. 05Fade the music out at the end.
  6. 06Clean up my audio and make my voice clearer.
  7. 07Add a subtle whoosh on each hard cut.
  8. 08Drop the music out completely during the important part.
  9. 09Add gentle ambient sound to the b-roll.
  10. 10Boost my voice, it's too quiet.

B-roll and visuals

  1. 01Add b-roll of a city skyline here.
  2. 02Cut away to relevant footage when I describe the process.
  3. 03Add a clip of hands typing on a laptop over this line.
  4. 04Show a nature shot while I talk about slowing down.
  5. 05Add b-roll of a busy coffee shop here.
  6. 06Overlay a slow zoom on a photo of the product.
  7. 07Replace the boring wall behind me with something more interesting.
  8. 08Add a quick montage of the highlights at the end.

Emphasis and framing

  1. 01Zoom in when I say the price.
  2. 02Punch in slightly whenever I make a key point.
  3. 03Reframe this vertical, keeping me centred.
  4. 04Add a subtle shake on the big reveal.
  5. 05Slow this moment down for emphasis.
  6. 06Crop in tighter on my face during the emotional part.
  7. 07Add an arrow pointing at the button I'm talking about.

Format, length and cleanup

  1. 01Make this vertical for TikTok.
  2. 02Make a 30-second version for Reels from the best moments.
  3. 03Get this under two minutes for LinkedIn.
  4. 04Export a 16:9 version for YouTube and a 9:16 version for Shorts.
  5. 05Blur the logo in the bottom corner (I'll circle it on the frame).

Why these work

The list is less useful than the patterns underneath it. Five things separate an instruction that lands from one that misses.

They name an outcome, not an operation

Compare "add a transition" with "make the jump between these two points less abrupt." The first names a tool and leaves the goal implicit. The second names the goal and leaves the tool open, which means there is something to aim at, and something to judge the result against.

Almost every prompt above is phrased as a desired state rather than a menu item.

They anchor to content, not timecode

"Zoom in when I say the price" is findable. "Zoom in at 01:47" requires you to find 01:47 first, and you will get it slightly wrong.

You always know what you said. Use it. This is the single highest-value habit on the list.

They are relative where precision does not exist

"Bigger", "quieter", "snappier", "warmer". You do not have an absolute vocabulary for caption size, and neither does anyone else. Relative adjustments against what you are looking at converge on the right answer in two or three passes, which is faster than specifying a number you would have to guess.

They are specific where specificity matters

For b-roll especially, generic requests produce generic footage. "A person at a desk" gets you the same anonymous stock-looking shot everyone else has. "A close-up of hands scrubbing a timeline on a cluttered desk, late evening, screen light" gets you a shot.

The rule: be relative about *adjustments*, specific about *content*.

They do one thing when the thing is fiddly

Batch the structural work. Separate the fine-tuning. If you ask for five subtle changes at once and the result is 70% right, you cannot tell which instruction went wrong.

How to use the list well

Chain them, do not dump them. The strongest workflow runs top to bottom through the sections above: pacing, then format, then captions, music and b-roll, then length. Each layer is easier once the one before it is settled. The order matters more than the phrasing.

Adapt the specifics. These are templates. Swap in your moments ("when I say the price" becomes "when I mention the deadline"), your style ("lo-fi" becomes "cinematic"), your targets ("30 seconds" becomes "15 seconds").

Point when words fall short. For anything tied to a location on screen (a logo, a blemish, an object to remove) pause on the frame, circle it, describe the change. Prompt 50 is the archetype: circle the logo, then say "blur this." How pointing works.

Treat it as a conversation. You do not need the perfect prompt. Make an edit, react, refine. "Bigger." "Quieter." "Put that back." The back-and-forth *is* the workflow, and it is faster than any timeline.

The prompts that do not work

Worth knowing the failure modes too.

  • "Make it better." No target. Better how, tighter, brighter, funnier, shorter?
  • "Make it go viral." Not an edit. It is an outcome nobody controls.
  • "Fix the audio." Fix what: the level, the noise, the room, the balance against the music? Name the symptom.
  • "Add some b-roll." Where, and of what? This produces generic filler, which is worse than none.
  • Long compound instructions. Six changes in one paragraph gives you one result and no way to isolate what went wrong.

The through-line: an instruction needs a target the result can be checked against. If you cannot say how you would know it worked, the instruction is not ready yet.

Try it for yourself

Describe the video you want, or upload a clip and describe the edit. No timeline, no learning curve, start free.

All posts