Free Video Effect
Vertigo Effect: A Dolly Zoom on Your Talking-Head Video
A dolly zoom: you stay the same size while the room behind you stretches towards you, with a bold word on screen.
Claude Code · Opus 5.59:16Advanced
- Run the setup once
- Paste the prompt, fill in the [brackets]
- Approve the preview, get your MP4
New to Claude Code? Start with the beginner guide →
Want all 17 at once? Get the full pack →
The prompt
effects/vertigo.py: add a "Vertigo" (dolly zoom) effect between [START S] and [END S].
- Separate me and the background with the person mask.
- In the background, fill the hole left by removing me from the surrounding pixels
(OpenCV inpaint, or temporally from earlier frames).
- During the effect:
- Background: grows from scale 1.0 → [1.35] (centered on my face), with a light
perspective distortion at the same time (the edges stretch more, the center less)
- Me: scale 1.0 → 0.97, so almost unchanged
- Easing: easeInOutSine, [1.2] s in total
- Add a very light radial blur to the background (stronger at the edges).
- The accent word "[WORD]" appears in the middle at the same time, in large, bold, white letters.
- You can also do it in reverse if you like: the background shrinks (1.35 → 1.0), which feels like "moving away".Fill in the brackets
| Placeholder | What to put there | Example |
|---|---|---|
[START S] / [END S] | Where the effect starts and ends, in seconds from the start of input.mp4. | 8.4 / 9.8 |
[1.35] | End scale of the background. 1.0 means no change; a higher number gives a stronger stretch but reveals more of the filled-in area behind you. | 1.35 |
[1.2] | Length of the background move, in seconds. | 1.2 |
[WORD] | The accent word shown in big, bold, white letters in the centre as the move begins. | WAIT |
About Vertigo
Vertigo is a dolly zoom effect for talking-head video: you stay almost the same size while the background behind you scales up and stretches around your face, so the room seems to lurch. After the one-time setup, paste the prompt below into Claude Code with Opus 5.5, and Claude separates you from the background, fills the gap and animates both layers.
Use it when: You share a surprising fact, reach a dramatic beat, or get to the line where you say "wait, this matters".
What the Vertigo effect does
Vertigo imitates the dolly zoom, the shot where a person holds their size in the frame while the space around them seems to warp. In this version you scale from 1.0 to only 0.97, while the background grows from 1.0 to 1.35 around your face and stretches more at the edges than in the middle. The move takes about 1.2 seconds on an easeInOutSine curve, and a large, bold white word appears at the centre as it starts.
The technique is commonly associated with Alfred Hitchcock’s Vertigo (1958), which is where the nickname comes from. Here it is rebuilt from a static shot, without moving a camera at all. The prompt also offers the reverse: the background shrinking from 1.35 back to 1.0, which feels more like the room pulling away from you.
When it works and when it hurts
It works on a single line that should land hard: a number nobody expected, a turn in a story, “stop, this part is important”. The disorientation is brief, so it reads as emphasis rather than as a scene change.
It hurts when it is overused or placed on a line that does not earn it. Viewers notice the warp, and on a routine sentence it can look like a glitch. It also needs something behind you to warp. Against a plain, evenly lit wall, a scale change has little to show and comes across as an ordinary zoom on the background.
How Claude builds it
Claude uses the person mask to split each frame into two layers: you, and the room without you. Lifting you out leaves a person-shaped hole in the room layer, so Claude fills it, either with OpenCV’s inpainting, which spreads surrounding pixels inwards, or by copying wall from earlier frames in which you had moved aside.
During the effect, the room layer is scaled up around the position of your face, with an extra perspective warp, and a radial blur is added that gets stronger towards the edges. Your layer goes back on top at nearly constant size. Because the fill sits behind your body, most of it stays hidden; it shows only where the growing background pushes it out past your outline, which is why the end scale matters. ffmpeg writes the result back to MP4 with your original audio.
Make it yours
The two numbers in brackets are the main controls: the background’s end scale and the duration. A smaller scale, around 1.15 to 1.2, gives a quiet unease; a larger one is more dramatic but exposes more of the fill. A shorter duration feels like a jolt, a longer one like a slow realisation. You can drop the centred word, set it in a serif instead of bold sans, or add a sound effect in your editor afterwards. Test one change at a time.
Questions
What is a dolly zoom?
It is a camera technique where the camera moves towards or away from the subject while the lens zooms the opposite way, so the subject keeps its size and the background seems to stretch or compress. Because of its use in Alfred Hitchcock's Vertigo (1958), it is often called the Vertigo effect.
Do I need to film it with a special camera move?
No. You film a normal, static talking-head shot. Claude splits you from the background with the person mask, scales and warps only the background, and keeps you almost still. The result imitates the look of a dolly zoom; it is not a real optical one.
Why does the background need filling in?
When Claude lifts you off the frame, a person-shaped hole is left in the background layer. As that layer grows past your outline, the hole would grow out from behind you and become visible. Filling it with nearby pixels, or with wall from earlier frames, keeps it hidden.
Which direction should I use?
Growing the background from 1.0 to 1.35 feels like the room closing in, which suits a shock or a reveal. Shrinking it from 1.35 to 1.0 feels like the space opening up or dropping away. Both are in the prompt; render each and pick the one that fits the line.
Step by step
Run the setup once
Open the folder that holds input.mp4 in Claude Code, choose Opus 5.5 and paste the setup prompt. Vertigo needs the person mask it creates to lift you off the background.
Find the line
Pick the one sentence that deserves the jolt. Note the second it starts and the word you want on screen.
Paste and fill in the prompt
Set the start and end seconds, the background scale, the duration and the word, and send it. Keep 1.35 and 1.2 for a first try so you have a baseline.
Review the edges of your body
Watch the area just around you as the background grows. If a smeared silhouette shows up, ask Claude to fill from earlier frames or to shorten the move, then re-render.
You need: Setup prompt (run once) · Person mask · Visible background around you · One accent word
Tips
- Frame yourself with some background showing: shelves, a doorway, the corners of the room. On a plain wall the scale change has little to show and reads as an ordinary zoom.
- Use it sparingly. It relies on surprise, so one per video is a sensible default; treat that as taste, not a rule.
- Try the reverse version (background shrinking from 1.35 to 1.0) when the line is about distance, loss or something slipping away. The prompt already includes it as an option.
- Keep the move short. The longer the background keeps growing, the more time viewers have to notice the filled-in area behind you.
If it goes wrong
A smeared ghost of your outline appears in the background as it grows.
That is the filled hole showing. Ask Claude to expand the mask by a few pixels before filling, prefer wall pixels from earlier frames over inpainting, and soften the filled area slightly.
A halo or flicker appears around your edges.
Ask Claude to feather the mask edge by 2 px, smooth it over time in that range and re-render only those seconds.
It looks like a plain zoom, not a dolly zoom.
Ask for a stronger perspective stretch at the edges, a slightly larger end scale and your own scale held closer to 1.0, so the two layers clearly move differently.
What it does not do
- It is a 2D fake. A real dolly zoom changes perspective because the camera physically moves; here the background is one flat image being scaled and warped, so near and far objects in your room do not separate.
- Filling the hole behind you is guesswork. OpenCV inpainting smears nearby pixels, and if you barely move, the wall behind you is never visible in any frame, so the temporal fill has nothing to copy. Patterned walls show the patch most.
- It depends on a clean person mask. Loose hair and fast hand movement show up as flicker or halos, because you and the background move differently at the edge.
- The extra scaling and warp soften the background slightly at higher end scales, since the frame is being enlarged beyond its original pixels.
Sources
- Dolly zoom — Wikipedia. Accessed October 21, 2026.
- Robust Video Matting (RVM) — GitHub — PeterL1n. Accessed October 21, 2026.
- Face detection guide — Google AI Edge — MediaPipe. Accessed October 21, 2026.
- Easing functions cheat sheet — easings.net. Accessed October 21, 2026.
Want the whole library?
Lifetime opens every AI Video Effect, AI Motion Graphic and AI Creator Concept, inside Claude via MCP, for one payment of $79.