20 AI video editor prompts for talking-head videos
Last updated 2026-10-20
An AI video editor prompt works best when it names the moment, the look and the length: where the edit happens, what it should look like and how long it lasts. Then you check the result and refine one thing. Below are 20 copy-paste prompts for talking-head videos, grouped by job, from cutting pauses to export.
How to write a good edit prompt
A prompt is a short brief for an editor who has not seen your video. Give it three things and the result lands close to what you pictured.
Woven's AI agent receives stills from your footage and reads speech as text with word timings. It cannot watch motion or hear sound. So point at words and times, not at feelings. Cut the dead air is vague. Cut every pause over half a second is something it can do.
- Say the moment. Give a time or quote the words, such as at 0:14 or on the word free. To point at one clip, use "Reference in chat" from its right-click menu.
- Say the look. Use names Woven shows, such as Punch, Highlight or Pill. A named look beats make it pop.
- Say the length. Give seconds: for 2 seconds, or about 30 seconds in total.
- Check it. Play the result. To see a caption position first, ask for a still of that moment.
- Refine one thing. Reply with a single change and leave the rest alone.
Prompts for cutting, punch-ins and captions
Cut first and caption last, because captions follow the final timing. Swap the times and words for your own.
| Job | Prompt | Use it when |
|---|---|---|
| Cutting | Transcribe the clip and cut every pause over half a second. Keep the pause before each punchline. | Dead air makes the video drag. |
| Cutting | Keep one take per line, the final one, delete my restarts and close the gaps. | You repeated lines while filming. |
| Cutting | Open the video on the sentence that states the result. Cut the greeting before it. | The opening wastes the first seconds. |
| Punch-ins | Add a Punch on the word free at 0:14 with the focal point on my face. | One word needs emphasis. |
| Punch-ins | Add a Push-in across the sentence that starts at 0:20, lasting the whole sentence. | A long sentence needs movement. |
| Captions | Generate captions in the Highlight look with the spoken word in yellow, placed at the bottom. | The cut is final. |
| Captions | Switch the captions to Word by word, in uppercase, centred on screen. | You want a fast, punchy read. |
| Captions | In the captions, change wovan to Woven and delete any um that shows up. | A name was misheard. Fixing words is free. |
Prompts for B-roll, text and titles
These cover what appears on screen besides your face. Generated clips use credits, so check pricing before you ask for several.
| Job | Prompt | Use it when |
|---|---|---|
| B-roll | Whenever I name a product, show a muted clip from my B-roll folder over that moment. | You name things you can show. |
| B-roll | Generate two short clips of a messy desk and place them between 0:08 and 0:14. | You have no footage of the thing. |
| B-roll | Shrink my screen recording into the bottom-right corner and keep my face full-size behind it. | You demo an app while you talk. |
| Titles | Add a title at 0:02 reading Three mistakes, in the Pill style, fading out after 2 seconds. | The hook needs a label. |
| Titles | Add my name and role along the bottom edge from the start to 0:03, in the Subtitle style, sliding up. | Viewers need to know who is talking. |
Prompts for sound, color, framing and export
Do these last, since each one depends on the final cut.
| Job | Prompt | Use it when |
|---|---|---|
| Music | Generate a calm instrumental track, set it to 15% volume beneath the entire video and let it fade as the video ends. | You want a bed, not a song. |
| Sound | Add mouse-click-2 at 0:31 and 0:36, where I click in the screen recording, at 40% volume. | A screen demo feels silent. |
| Sound | Set the music to 10% and each sound effect to 35% so my voice stays the loudest thing. | Sound competes with your voice. |
| Framing | Switch the canvas to Vertical 1080x1920 and fill the frame, centred on me. | The take is horizontal. |
| Color | Warm up the main clip: Warmth +20 and Contrast 110%. | The footage looks cold or flat. |
| Check | Render two stills, at 0:05 and 0:20, so I can check captions and framing before I export. | You want to check captions and framing. |
| Export | Export the reel as an MP4. | You are done. Chat exports are 1080p; for 1440p use Export in the toolbar. |
What a prompt will not do
Woven has no pause-remover button. The first three prompts work because the agent transcribes the clip, reads the word timings and cuts where you say. Transcribing and generating both spend credits. Speech Woven transcribed earlier is not charged again when you caption it.
It does not follow your face. A Punch keeps the focal point where it is set, so check the framing. It cannot hear the result and does not turn the music down under your voice by itself, so set the music volume low and check the sound by playing the reel.
Use these prompts in Woven
Paste a prompt into Woven's chat with your clip open on the timeline. The AI agent edits the open reel and you see each change land. Fixing caption words costs nothing. Transcription, captions for speech Woven has not transcribed and generation use credits.
Ask in chat
- Cut the pauses, then caption the result in the Highlight look.
- Show me stills at 0:05 and 0:20.
- Make that Punch shorter, about half a second.
Or do it by hand
- 1Open the clip. Use "Add" › "Video" for your clip, or drop it straight onto the timeline.
- 2Paste one prompt. Swap in the times and words from your own video, then send it.
- 3Point at what you mean. Right-click a clip and choose "Reference in chat", so the agent edits that exact item.
- 4Play it back. Press Space and watch. The agent sees stills only, so you judge the motion and the sound.
- 5Refine. Reply with one change, such as a shorter Punch or a lower music volume.
Common questions
Name the moment, the look and the length, for example: add a Punch on the word free at 0:14 for 1 second. Play the result, then change one thing at a time.
Yes, for footage you already have. Woven's AI agent cuts, captions, adds effects and music from chat. It sees stills and reads a transcript, so you check motion and sound yourself.
Specific ones: a time or quoted words, a named look and a length. The 20 prompts above follow that pattern for talking-head videos.
Woven has no automatic remover. Ask for it in chat: the agent transcribes the clip, then cuts the pauses and any filler words that show up in the transcript, using the word timings.
The prompt was probably missing a time, a named look or a length. Add the missing piece and ask again for just that change.
Try Woven free for 3 days
Then $25/mo billed yearly ($299/yr) or $39/month. Native Mac app for short-form video.
Paste into Claude Code to install and connect Woven.