How-to · intermediate · 3 min read
How to cut a talking head video so it feels crisp
Cut every pause over 0.3 seconds, every filler, every bad take. Then hide the jump cuts with punch-ins and b-roll, and leave a small breath before each idea.
By Sulan Zhang, co-founder of Vyra · Updated 2026-09-24 · Example: @sulansart
- Level
- intermediate
- Read time
- 3 min
A crisp talking head is a speech edit first. Remove pauses over about 0.3 seconds, fillers, false starts, and repeated takes so it sounds like the best version of you. Then cover the jump cuts by switching between the full shot and a 15 to 25 percent punch-in, or dropping b-roll on the cut. Leave a small breath before each new idea so it doesn't sound like a machine.
Why it matters
People forgive bad lighting.
The method
- Pick the best take per sentence. Long recordings have three versions of most lines. Keep the clearest one.
- Remove silence and filler. Pauses over 0.3 seconds, "um," "like," "so basically," any restart. Keep a 0.1 to 0.2 second breath before a new point.
- Hide each jump cut. Punch in, switch angle, insert b-roll, or land a caption emphasis on the cut. Rotate.
- Punch in on the important word. Not on the cut. 15 to 25 percent.
- Cover big removals with b-roll. Over 2 seconds cut, add a 1.5 to 3 second insert.
- Check the audio seams. Cut between words, never inside one. A 2 to 4 frame crossfade if it clicks.
Rules of thumb
| Decision | Number |
|---|---|
| Pause to remove | over 0.3 s |
| Breath to keep | 0.1 to 0.2 s |
| Punch-in | 15 to 25 percent |
| Punch-ins in a row | 2 max, then b-roll or angle |
| B-roll insert | 1.5 to 3 s |
Examples
- Raw, 11 seconds. "So, um, the thing about, like, pricing is, okay let me start over. The thing about pricing is that most people undercharge because they, uh, they're scared. They're scared of losing the client."
- Cut, 6 seconds. "The thing about pricing is that most people undercharge because they're scared of losing the client." Full shot on the first clause, punch-in on "undercharge," b-roll of an invoice on "losing the client."
Do it in Vyra
- "Cut this talking head down. Remove all pauses over 0.3 seconds, filler words, and any sentence I restart. Keep the best take of each line. Then hide the jump cuts by alternating a 20 percent punch-in and my b-roll clips. Keep a small breath before each new point."
- "Clean up the transcript first, then tighten the edit to match. Do not cut inside words. Add a 3 frame audio crossfade on every cut."
- "This edit sounds too rushed. Add back 0.2 seconds of breath before each new sentence and replace the punch-ins from 14 to 22 seconds with the wide angle."
Common mistakes
- Cutting every pause to zero. It sounds robotic.
- A punch-in on every cut. People notice by the fourth.
- Cutting inside a word. Always audible.
FAQ
How much will I cut? A 10 minute recording usually becomes 3 to 5 minutes. For short-form, 8 minutes becomes 45 to 60 seconds.
Can I do it without b-roll? Yes. Two angles, or punch-ins plus caption emphasis.
Related
- /how-to/choose-captions
- /how-to/frame-yourself
- /how-to/shoot-b-roll
- /formats/talking-head