How to blur the background of a video
To blur the background of a video while keeping the subject sharp, you separate the subject from the background using a mask, then apply a blur filter to everything outside that mask. The mask is the hard part: it has to follow the subject frame by frame without flickering at the edges.
There are three ways to get that mask, and the choice determines quality and effort:
1. Shoot shallow depth of field
Optical blur (bokeh) from a wide-aperture lens. Zero post work. No flicker. Best quality. Requires a real camera with manual aperture control, not a phone in auto mode.
2. AI subject segmentation
Software detects the subject and generates a mask per frame, then blurs outside it. Works on existing footage. Can flicker at edges without temporal smoothing.
3. Manual mask + track
Draw a mask around the subject, track it through the clip, feather the edge, blur outside. Full control. Labour-intensive on moving subjects. Works in any editor.
Green screen shortcut
If you shot on green screen, key the green first to get a perfect matte, then blur the background layer separately. Cleanest edge, no flicker, deterministic mask.
Blurring is not removing. Background removal deletes the background; transparent video makes it absent (alpha = 0). Blur keeps the background visible but defocused. If you want the subject on a new background, you need removal, not blur.
Blurring a video background is a compositing operation: you isolate the subject on one layer, blur the background on another layer, and combine them. The subject layer must have a mask (or matte) that separates it from the background. Without that separation, the blur applies to the entire frame, including the subject, which is not what anyone means by "blur background video."
Three operations that are not the same
People conflate three different background operations. The tool, the export format, and the result are different for each:
| Operation | What happens to background | Requires mask? | Export format |
|---|---|---|---|
| Blur background | Visible but defocused | Yes (subject mask) | Standard MP4 (H.264) |
| Remove background | Deleted, replaced by new image or colour | Yes (key or mask) | Standard MP4 (composited) |
| Make transparent | Absent (alpha = 0) | Yes (key or mask) | WebM alpha / ProRes 4444 / PNG seq |
Blur produces a normal video file with no alpha channel. The background is still there, just out of focus. This means you can deliver it as a standard MP4, and every player and platform supports it. The trade-off is that the blur is baked in: you cannot un-blur it later.
Method 1: Shoot shallow depth of field (optical bokeh)
The best blur background video is the one you do not have to do in post. Shooting with a wide aperture (f/1.4, f/1.8, f/2.8 depending on the sensor size) produces a natural depth-of-field blur called bokeh. The background is optically defocused at the moment of capture, the subject is sharp, and the result requires zero editing.
Optical bokeh has three properties that no software blur can match: it is temporally stable (no flicker between frames), it is physically accurate (the blur follows the depth map of the real scene, so distant objects blur more than near ones), and it has no mask edge artifacts (there is no computed boundary, the transition is continuous).
To get it: use a camera with a large sensor (full-frame or APS-C) and a fast lens (f/1.4 to f/2.8). Set the aperture wide open. Increase the distance between subject and background. The further the background is from the subject, the more it blurs. A 50mm lens at f/1.8 on a full-frame body, with the subject 3 metres from the camera and the background 8 metres behind the subject, produces a strongly blurred background.
Phone cameras simulate this with Portrait mode, which uses depth sensing (LiDAR on recent iPhones, dual-camera parallax on older models) to compute a depth map and blur the background algorithmically. The result is close to optical bokeh but can show edge artifacts around hair and fine details, and the blur radius is uniform rather than depth-graded. For professional video, optical capture is the reference.
Method 2: AI subject segmentation in post
If you already have footage with a sharp background, AI segmentation can generate a subject mask per frame, and you blur everything outside that mask. This is the most common approach for existing clips, because it does not require a green screen and works on footage shot in any environment.
DaVinci Resolve Studio (paid): On the Color page, add a node, open the Qualifier panel, and use the Smart Selector or the Neural Engine person mask. The Neural Engine generates a subject mask using on-device inference. Invert the mask so the background is selected, then apply a Blur (Gaussian or Lens Blur) to the inverted selection. For temporal stability, enable Object Mask tracking instead of per-frame detection, so the mask follows the subject across frames rather than being regenerated each frame.
Adobe Premiere Pro: Use the Masking tool on an adjustment layer. Draw a mask around the subject, set it to invert (so it affects the background), and apply Gaussian Blur. Track the mask forward. For better subject isolation, use the Auto Reframe feature to keep the subject centred, or use the Selective Color key to refine the mask edge. Premiere's mask tracking is less accurate than Resolve's Neural Engine on complex motion, so expect to keyframe manual adjustments on busy clips.
CapCut (free, desktop and mobile): Import the clip, go to Effects, and apply a Blur or Bokeh effect. Use the mask tool to protect the subject area. CapCut's automatic subject detection works for simple shots (single person, clear contrast against background) but struggles with multiple subjects or busy backgrounds. For those cases, switch to manual masking.
After Effects: Use the Roto Brush tool to generate a subject matte. Roto Brush propagates the matte across frames using edge detection and motion estimation. Apply Gaussian Blur (or Camera Lens Blur for a more realistic bokeh) to a duplicate layer beneath the subject matte. Feather the matte edge by 2 to 5 pixels to avoid a hard boundary. For complex shots, use the Mocha AE tracker for planar tracking of the mask.
FFmpeg (command line): FFmpeg does not do AI segmentation. It can apply a uniform box blur or Gaussian blur to the entire frame, or to a rectangular region. For subject-aware blur, you need a separate tool to generate the mask (a PNG matte per frame), then use FFmpeg to apply the blur through that matte. The command for a simple uniform background blur on a rectangular region: ffmpeg -i input.mp4 -filter:v "boxblur=enable='between(t,0,10)':luma_radius=20" -c:a copy output.mp4. This blurs the entire frame, not just the background, and is only useful when you then overlay the sharp subject from a separate layer.
Method 3: Manual mask and track
When AI segmentation fails (complex backgrounds, multiple subjects, low contrast), manual masking is the fallback. You draw a mask around the subject, feather the edge so the transition is not a hard line, and track the mask through the clip so it follows the subject's movement.
The workflow in any editor that supports masks (Resolve, Premiere, After Effects, Final Cut Pro):
- Duplicate the clip onto two layers: subject layer on top, background layer on bottom.
- Draw a mask on the subject layer around the subject. Use a shape mask (ellipse for simple shots, polygon for complex outlines).
- Feather the mask edge by 3 to 8 pixels. This softens the boundary so the blur does not start at a hard line.
- Track the mask forward through the clip. If the track drifts, keyframe manual position adjustments at the drift points.
- Apply blur to the background layer (the bottom layer). Gaussian Blur for a smooth look, Camera Lens Blur or Iris Blur for a bokeh-like appearance.
- Adjust blur radius to taste. Start at 20 to 40 pixels for 1080p footage, less for close-ups, more for wide shots.
- Render and review at full resolution. The preview at half resolution may hide edge artifacts.
Manual masking takes 5 to 30 minutes per minute of footage depending on subject motion. For talking-head videos where the subject barely moves, it is fast. For action clips with rapid movement, it is slow and AI segmentation or a green screen reshoot is more practical.
The green screen path to a clean blur
If you are shooting specifically for a blurred background and have control over the set, using a green screen gives you the cleanest result with the least post-production effort. Key the green to produce a matte, place the subject on one layer and the original background (or a new one) on a second layer, blur the background layer, and composite. The matte is deterministic, the edge is clean, and there is no flicker because the key is stable across every frame.
The advantage over AI segmentation: the mask edge is exact (it is a colour key, not an inference), so hair, motion blur, and semi-transparent areas are handled by the keyer's spill and despill controls rather than approximated by a neural network. The disadvantage: you need a green screen and even lighting on it, which is not available in run-and-gun or outdoor shoots.
For the blur step after keying, apply the blur to the background layer only. The subject layer (the keyed foreground) stays sharp. Because the two layers are separate, you can adjust the blur radius at any time without affecting the subject edge. This is the same compositing logic used for background replacement, except instead of swapping the background image, you blur the existing one.
Why per-frame AI blur flickers and how to reduce it
AI segmentation generates a mask for each frame independently. When the subject moves, the mask boundary shifts slightly between frames, and pixels near the edge alternate between blurred (outside the mask) and sharp (inside the mask). The viewer perceives this as a shimmering or flickering halo around the subject, which is more distracting than no blur at all.
Three ways to reduce it:
1. Temporal smoothing. Use the editor's object tracking or temporal smoothing option instead of per-frame detection. DaVinci Resolve's Neural Engine Object Mask tracks the mask across frames rather than regenerating it, which stabilises the boundary. After Effects' Roto Brush propagates the matte forward with temporal interpolation.
2. Wider feather. Increase the mask feather to 8 to 15 pixels. A wider feather creates a gradual transition from sharp to blurred, so small mask shifts produce a smaller visible change. The trade-off is that the subject edge becomes softer, which may look slightly out of focus.
3. Lower blur radius near the edge. Instead of a single blur value, use a gradient: full blur at the frame edges, reducing to zero blur near the subject. This is done with a vignette mask or a radial gradient on the blur filter. Pixels near the subject have low blur, so mask shifts produce minimal visible change. This sacrifices some background blur near the subject but eliminates the flicker.
Blur on phone footage
Phone cameras in Portrait video mode apply a computational blur at capture time. This produces a result that looks like shallow depth of field, but the blur is based on a depth map computed by the phone, not optical defocus. The quality depends on the phone's depth sensing: iPhones with LiDAR produce more accurate depth maps than phones using only dual-camera parallax.
For clips not shot in Portrait mode, the options are: CapCut's blur effect (free, works on both platforms), or a desktop editor with AI segmentation (Resolve, Premiere). CapCut's automatic subject detection works for single-person shots with clear background separation. For complex shots, export the clip to a desktop editor.
Phone Portrait mode has a fixed blur radius set at capture. You cannot increase or decrease it in post, and the edge artifacts (visible around hair and transparent objects) are baked in. If you need adjustable blur, shoot in normal mode and apply the blur in post.
Choosing a blur type
Editors offer several blur filters. The visual difference is in how the blurred pixels are rendered:
| Blur type | Appearance | Use when |
|---|---|---|
| Gaussian Blur | Smooth, uniform softening | General purpose, privacy blur, hiding detail |
| Camera Lens Blur / Iris Blur | Bokeh-like, brighter highlights retain shape | Simulating shallow depth of field, cinematic look |
| Box Blur | Blocky, less smooth than Gaussian | Fast preview, low processing cost |
| Tilt-Shift / Depth Map Blur | Graduated blur based on depth or position | Simulating depth without a real depth map |
Gaussian is the default in most editors and works for most use cases. Camera Lens Blur (called Iris Blur in some editors) produces a more realistic bokeh because it preserves the shape of bright highlights, which Gaussian blur flattens. For a cinematic blur background video, Camera Lens Blur is the better choice if your editor offers it and your hardware can render it in reasonable time.
Export settings for blurred video
A blurred background video is a normal composited video file. There is no alpha channel, no special format, no transparency. Export as MP4 with H.264 encoding for general delivery, or H.265 for smaller file sizes. The blur is baked into the pixels, so every player and platform renders it identically.
Set the bitrate based on the resolution: 8 to 12 Mbps for 1080p, 20 to 35 Mbps for 4K. The blurred background compresses well because it has less high-frequency detail, so you can use a slightly lower bitrate than a fully sharp video of the same resolution without visible quality loss in the subject.
Do not export as transparent video (WebM with alpha or ProRes 4444). The blur is not transparency. If you need the subject on a transparent background for compositing elsewhere, that is a background removal operation, not a blur operation, and requires a different export format.
Quick answers
How do I blur the background of a video on my phone?
On iPhone, open the clip in Photos and use the Portrait mode blur slider if the clip was shot in Portrait mode. For regular clips, use CapCut: import the video, tap Effects, choose Blur or Bokeh, and use the mask tool to protect the subject. On Android, CapCut offers the same workflow. For existing non-Portrait clips, the blur is applied uniformly unless you mask the subject area, which means the subject also gets blurred unless you cut a mask around it frame by frame.
What is the difference between blurring the background and removing the background?
Blurring the background keeps the background visible but out of focus, so you see a soft, defocused version of the original scene behind the subject. Removing the background deletes it entirely, replacing it with transparency (alpha channel) or a new solid colour or image. Blurring is a focus effect; removal is a keying or masking operation. You blur when the background is part of the story but should not distract; you remove when you want to composite the subject onto something else.
Can I blur the background of a video without a green screen?
Yes. You can use AI portrait segmentation to generate a subject mask, then apply a blur to everything outside the mask. DaVinci Resolve Studio has a built-in Neural Engine person mask. Premiere Pro supports auto-reframe and masking. CapCut has an automatic subject detection in its blur effect. Without AI, you can manually draw a mask around the subject and animate it frame by frame, which is labour-intensive but works in any editor.
Why does my blurred video background flicker?
Flicker happens when the mask boundary shifts between frames, causing pixels near the subject edge to alternate between blurred and sharp. This is most visible with AI segmentation on footage where the subject and background have similar colours. To reduce it: widen the mask feather so the transition is gradual, track the mask to the subject instead of using per-frame detection, or shoot against a green screen so the mask is deterministic. Per-frame AI masks without temporal smoothing produce the worst flicker.
Does shooting with a shallow depth of field blur the background automatically?
Yes. Shooting with a wide aperture (low f-number, such as f/1.4 or f/1.8) on a real camera produces a natural background blur called bokeh. This blur is baked into the footage, requires no post-processing, and does not flicker because it is optical, not computed. Phone cameras simulate this with Portrait mode, which uses depth sensing to blur the background computationally. The simulation can show edge artifacts, especially around hair, that optical bokeh does not produce.
Related reading: removing a green screen from video (the keying path to a clean mask before blurring), removing a video background without a green screen (when you want the background gone, not blurred), transparent background video formats (when you need alpha, not blur), and choosing a green screen editor (which editors support masking and tracking for blur).