Skip to contentviewgum

How to edit a talking-head take for Shorts, TikTok and Reels

Start on the first word, cut pauses longer than 0.35 seconds along with false starts and filler words, change the framing in small steps instead of zooming on every cut, and burn in captions 2 or 3 words at a time at mid-chest. Put a picture or graphic tied to the words on screen every few seconds, and end a fraction of a second after the last word.

Updated October 18, 20265 min read

Film for the cut

  • Hold the phone upright and fill the frame from mid-chest up. A vertical original keeps its sharpness; a horizontal take cropped to vertical loses most of its pixels.
  • Put your eyes about a quarter to two fifths of the way down the frame (y 500 to 800 on 1080 x 1920), so captions fit under the chin.
  • Leave a clear beat of silence between attempts at a line. It makes the cut points obvious.
  • Say the point first if you can. If the take opens with context, the edit can move the line it picks to open to the front.

Cut the silence, keep the breath

  • The first frame lands on the first syllable: no inhale, no fade from black, no logo.
  • Cut pauses longer than about 0.35 seconds. Calm, explanatory niches can keep a little more.
  • Cut filler words (um, uh, a filler "like"), false starts and repeated attempts. Keep the last complete attempt unless an earlier one is clearly better.
  • Keep 0.3 to 0.6 seconds of silence before a punchline or reveal. A pause that sets up a point is part of the point.
  • Cut greetings, channel intros and lines like "like and subscribe". They give a new viewer nothing.

Framing and punch-ins

  • Jump cuts hide better with a small change of framing on each one, in varied steps. Strict alternation between two zoom levels reads as a template.
  • Save a strong punch-in (1.25x to 1.4x) for the peak word of a beat, at most once every 10 seconds.
  • Re-centre on the face after every scale change so the eyes stay on the same line.
  • No constant zoom pulsing and no shake on every word.

Captions

Captions carry the video for anyone watching muted. In a 2019 survey of 5,616 US adults, 69% said they watch video with the sound off in public (Verizon Media and Publicis Media, reported by Streaming Media).

  • Show 2 or 3 words at a time, broken on phrase boundaries. Never split a number from its unit or a name across two captions.
  • Highlight at most one word or number per caption, and pick words that carry the meaning.
  • Sentence case, at mid-chest, with a stroke or shadow for contrast. Never across the mouth.
  • Keep them inside the safe zones, clear of each app's buttons and post text.
  • Check every name and number against what was said. A wrong caption on a talking head undermines the speaker.
  • TikTok's ad guidance suggests 5 to 10 words a second for on-screen text (TikTok Ads Help). Treat that as a ceiling for captions that echo speech, not as a pace for cards a viewer must read on their own.

Graphics, cut-ins and b-roll

  • Change something small on screen every 2 to 4 seconds (a reframe, a label, a highlight) and something bigger every 5 to 8 seconds (a cut-in, a graphic, a new picture).
  • Show the specific thing being said: this invoice, this screen, this machine. A generic image for a general idea reads as decoration.
  • Bring a graphic in on the word it belongs to, hold it long enough to read, and take it off when the point is done.
  • At most two elements on screen besides the captions, and never one over the face.

Sound

  • The voice comes first. Advice and education videos often carry no music under speech; if they do, keep it well under the voice.
  • Match the music to the claim. A serious point under a playful track loses credibility.
  • Use sound effects only on something that happens on screen, no more than one every 2.5 seconds and about 3 a minute.

End and export

  • End 6 to 12 frames (0.2 to 0.4 seconds at 30 frames a second) after the last word, unless a graphic still needs reading time.
  • Export a full-bleed 1080 x 1920 MP4 with no borders and no other app's watermark. Instagram makes watermarked or bordered reels less visible (Instagram).
  • Write the post copy for each platform. The specs guide has the limits.

What reads as low-effort

  • Auto-captions with wrong names, no emphasis, or placement over the face or the buttons.
  • A zoom on every cut, every word animated, or a sound effect on every cut.
  • A cut, a punch-in, a sound and a graphic all landing on the same word.
  • One static shot for 45 seconds with nothing else moving, or the opposite: cuts with no reason.
  • Music louder than the voice.

Where ViewGum fits

Upload one take filmed upright on your phone. ViewGum picks an opening line and moves it to the front, cuts the pauses, false starts and filler words, frames it for vertical, and adds captions, graphics and sound. Ask for changes in plain words or pick another style, then export for each platform.

Questions

How long a pause should I cut?

Anything over about 0.35 seconds, except the 0.3 to 0.6 seconds before a punchline or reveal.

Should captions show one word at a time?

It works for fast, high-energy videos. For advice and explainers, 2 or 3 words at a time in sentence case is easier to follow.

Do I need b-roll?

Not always, but a single uncut shot of a face for a whole minute gives the eye nothing new. A graphic, label or cut-in every few seconds, tied to the words, does the same job as b-roll.

Sources

All checked October 18, 2026

  1. Streaming Media: 80% of video caption users aren't hearing impaired, finds Verizon (May 20, 2019)
  2. TikTok Ads Help: Creative best practices for performance ads (updated June 2025)
  3. Instagram: Instagram Ranking Explained (May 31, 2023)

Edit your first take

Upload a take, get the first cut, change what you like, and export it for every platform.

Start free

Sign in with Google. New accounts get bonus credits to start.