# How to Blur a Face in DaVinci Resolve: Every Method > **Quick answer:** Blur a face in DaVinci Resolve by drawing a Power Window around it on the Color page, tracking it through the clip with the Cloud Tracker or IntelliTrack, then adding Mosaic Blur or Gaussian Blur from Resolve FX Blur to that same node. Studio adds Magic Mask for a tighter, hands-free matte on moving subjects. *Published by [TryUncle](https://tryuncle.com) — the on-screen assistant that teaches DaVinci Resolve on your own screen.* *Updated 2026-07-24 · DaVinci Resolve 21.0 (June 2026) · Canonical: https://tryuncle.com/learn/davinci-resolve/how-to-blur-a-face-in-davinci-resolve* Someone in your interview didn't sign a release. A bystander walks through your street footage. A minor shows up in a clip you're cutting for a school project. You need that face gone, but you don't want to crop the whole shot or reshoot it. DaVinci Resolve can blur it, track it, and keep it hidden for the full length of the clip, and every tool involved is available whether you're on the free version or Studio. We've spent seven years cutting commercial video in Resolve and built a video-editing community that's now past 100,000 members, and this exact question, "how do I blur a face without it sliding off," comes up constantly among people just past the beginner stage. It's a five-minute job on a static shot and a genuinely fiddly one on a shot where the person keeps turning their head. This guide covers both, plus the Studio shortcuts, the Fusion-page alternative, and the honest limits of what a blur actually hides. ## How do you blur a face in DaVinci Resolve? Blurring a face in DaVinci Resolve takes three ingredients on the Color page: a Power Window shaped over the face, a track that carries that window with the face as it moves, and a blur filter from Resolve FX Blur applied to the same node. All three are available in the free version, no Studio license required. Here's the shape of the job before you touch a single tool: 1. **Select your clip** on the Color page and add a serial node. 2. **Draw a Power Window**, usually a circle or ellipse, over the face. 3. **Track the window** through the clip if the face or the camera moves, using the Cloud Tracker or, on Studio, IntelliTrack. 4. **Apply a blur filter**, most often Mosaic Blur, to that same node. 5. **Raise the blur strength** until the face is unrecognizable, not just soft. 6. **Check the full clip at full resolution** before you export, since a tracked blur can hide gaps at reduced preview quality. That's the entire workflow for one static or slowly moving face. Where it gets more involved is a face that turns in profile, gets briefly blocked by a hand or another person, or shares the frame with several other faces you also need hidden. The rest of this guide walks through every one of those cases, plus the Studio shortcuts (Magic Mask, IntelliTrack) that trade manual tracking time for GPU processing, and the Fusion-page path that does the identical job with a Blur node instead of a Power Window. **A face blur in DaVinci Resolve is really just a tracked mask feeding a blur filter, and every version of that workflow, free or Studio, comes down to the same two questions: does the face move, and how many faces are there.** Answer those two questions honestly before you open a single effect, and the rest of this guide turns into a lookup table instead of a guessing game. ## Is the face moving or staying still, and why does that decide your whole workflow? Before you draw a single window, answer one question: does the face stay in roughly the same position and size for the entire shot, or does the person move, turn their head, or does the camera move relative to them? This single answer eliminates most of the wrong turns people take on this task. Scrub through the clip from the first frame the face needs to be hidden to the last. Watch for three separate kinds of motion, because each one stresses the tracker differently: - **Position and size.** Does the person walk, lean, or does the camera pan, zoom, or dolly? This is the baseline motion every tracker handles first. - **Head rotation.** Does the person turn toward or away from camera, tilt their head, or nod? A face turning from full-front to profile presents the tracker with a genuinely different set of pixels partway through the shot. - **Occlusion.** Does a hand, a microphone, another person, or a doorframe cross in front of the face at any point? If the answer to all three is no, a static Power Window and a single blur pass is a two-minute job regardless of which version of Resolve you're running. If the face moves but doesn't rotate or get blocked, the standard Cloud Tracker handles it cleanly. If the head rotates through a meaningful range or gets briefly occluded, that's where a plain Power Window track starts to struggle, and Studio's IntelliTrack or Magic Mask genuinely earns its keep. | Face motion | Head rotation | Occlusion | Recommended tool | | --- | --- | --- | --- | | None | None | None | Power Window, no tracking needed | | Position/size only | Minimal | None | Power Window with Cloud Tracker | | Position/size | Turns to profile or away | Occasional | IntelliTrack (Studio) or Magic Mask (Studio) | | Position/size | Turns freely | Frequent, other people cross in front | Magic Mask with manual correction, or segment the shot | This is the exact same diagnostic step that matters for [DaVinci Resolve's Power Window tracker in general](https://tryuncle.com/learn/davinci-resolve/davinci-resolve-power-window-keeps-losing-tracking), because a face blur is just a Power Window doing a specific job. A tracker that drifts off a face behaves identically to a tracker that drifts off any other tracked subject, and the fixes are the same: check contrast, check which transforms are enabled, and diagnose drift versus a hard break before touching a setting. **Whether a face rotates or gets briefly blocked matters more to your tool choice than whether it's a close-up or a wide shot, so diagnose motion first and pick your tracker second.** A wide shot of someone standing still is an easy job even in a crowded frame. A tight close-up of someone shaking their head no is a hard job even with nothing else happening around them. ## How do you blur a static face with a Power Window and Mosaic Blur? This is the fastest path in this entire guide, and it works identically whether you're on the free version or Studio, because none of the tools it uses touch the Neural Engine. 1. **Go to the Color page** and select the clip with the face you want hidden. 2. **Add a serial node.** Right-click the current node and choose Add Node, Add Serial, so your blur lives on its own step in the pipeline rather than mixed into your grade. 3. **Open the Window palette** and choose the Circle or Ellipse Power Window from the toolbar. 4. **Position and size the window** directly over the face, leaving a small margin so the blur fully covers the edges of the head as well as the face itself. 5. **Open the Effects Library**, search for Mosaic, and drag Mosaic Blur onto the node carrying your Power Window. 6. **Open the Mosaic Blur settings tab** and raise Pixel Frequency until the face is genuinely unrecognizable, not just softened. Per [Blackmagic's own DaVinci Resolve manual](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part3161.htm), Mosaic Blur is described plainly as **"a simple, pixelated blur suitable for hiding the face of anonymous witnesses."** That's not marketing language from a tutorial site, it's Blackmagic's own stated purpose for the filter, and it tells you exactly which effect to reach for first on this task. Beyond Pixel Frequency, Mosaic Blur gives you a handful of other controls worth knowing: - **Cell Shape** switches the mosaic blocks between Square, Hexagon, and Triangle. Square is the standard, most recognizable "pixelated" look. - **Aliasing** adjusts how tightly each cell samples the area underneath it. - **X Offset and Y Offset** shift the grid position, useful if a hard edge in the pattern happens to land right across a distinguishing feature like an eye. - **Antialias Result** softens the boundaries between cells slightly, trading a little bit of pure block-edge sharpness for a marginally less harsh look. - **Whole Cells Only** restricts the pattern to cells that sit entirely inside your window, which keeps the mosaic from producing half-cut, oddly shaped blocks right at the window's edge. Per Boris FX's own walkthrough of this exact workflow, author [Marco Sebastiano Alessi lays out the same core steps](https://borisfx.com/blog/how-to-blur-part-of-a-video-in-davinci-resolve-in/): draw the Power Window, track it if needed, then "select the blur tool and adjust the Radius sliders" or pull a named filter like Mosaic Blur from the Effects library and adjust it in the Settings tab. Two independent sources, Blackmagic's own manual and a plugin vendor with no reason to favor one Resolve-native tool over another, land on the identical recommendation. **Mosaic Blur is Blackmagic's own suggested tool for hiding a face, described in its manual using almost exactly that phrase, so it's the right default rather than a workaround.** Reach for it first. Reach for Gaussian Blur or a different filter only when Mosaic Blur's blocky look doesn't fit the tone of your project, covered in full a few sections down. ## How do you track and blur a face that moves through the shot? Once your motion check from earlier shows the face or the camera moving, a static Power Window will slide off target within a few seconds. This is where tracking comes in, and DaVinci Resolve gives you two different tracking engines to do it, depending on your license. **The Cloud Tracker, free in every version of Resolve**, is the default tracking mode in the Window Tracker palette. Per [Blackmagic's manual on Motion Tracking Windows](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part2864.htm), it works by analyzing "a cloud of tracking points that follow the vectors of every trackable group of pixels within the window you've created." Here's the workflow: 1. **Draw your Power Window** over the face on its clearest, most stable frame, not necessarily frame one of the clip. 2. **Open the Tracker palette**, next to the Window palette in the same toolbar. 3. **Check only the transforms the face genuinely needs.** For a face that's turning or nodding but not moving toward or away from camera, leave Pan and Tilt checked and uncheck Zoom, Rotate, and Perspective 3D to start. Add them back individually only if the track drifts without them. 4. **Click Track Forward and Back** from your anchor frame, so the tracker works outward in both directions from your best, clearest data rather than starting from a frame where the face might be blurred by motion or partially out of frame. 5. **Apply Mosaic Blur** to the same node once the track holds through the full range the face needs to stay hidden. **Studio's IntelliTrack, a DaVinci Neural Engine feature introduced in DaVinci Resolve 19, is the sturdier option for a face that turns, nods, or gets briefly blocked.** Per [RedShark News' in-depth writeup by Blake Jones](https://www.redsharknews.com/in-depth-with-davinci-resolve-19s-intellitrack-and-colorslice-features), IntelliTrack "is trained on real world examples," which makes it "less likely to fail in scenarios like tracking a subject behind brief occlusions" compared to the classic Cloud Tracker. Jones documented the standard tracker losing a subject that briefly passed behind a tree, a near-identical failure mode to a face being blocked by a hand or a second person, while IntelliTrack held the track through the same interruption. Jones sums up the appeal directly: **"What makes IntelliTrack so powerful is that everything is essentially done in the background."** To use it, open the Tracking area on the Color page, switch the tracking mode from Cloud Tracker to IntelliTrack before you add your tracking point, and track as normal. It's available in the Color page tracker, the Fusion point tracker, and the FX tracker, but only in DaVinci Resolve Studio. | Tracker | License | Best for a face | Weakness | | --- | --- | --- | --- | | Cloud Tracker | Free | Position, size, and mild rotation on a clear, high-contrast face | Drifts on low contrast skin, struggles through occlusion | | IntelliTrack | Studio only | Rotation, brief occlusion, busy or cluttered backgrounds | Requires Studio license and more GPU headroom | | Magic Mask | Studio only | Irregular motion, hair, profile turns, when you want the mask to hug the actual head shape | Heaviest GPU cost of the three, one Neural Engine pass per face | **A face that only shifts position tracks fine with the free Cloud Tracker, but a face that turns, nods, or gets briefly blocked is exactly the scenario IntelliTrack was built to handle, and it's worth the Studio upgrade if this is a recurring job for you.** If you're chasing a drifting track on a face specifically, the underlying causes and fixes are identical to any other tracked subject, and our guide to [Power Windows that keep losing tracking](https://tryuncle.com/learn/davinci-resolve/davinci-resolve-power-window-keeps-losing-tracking) covers the full diagnostic process, including the Clip mode versus Frame mode distinction you'll need once you're hand-correcting a stubborn frame or two. ## Mosaic Blur vs Gaussian Blur vs Box Blur: which one actually hides a face? Resolve FX Blur ships with several filters that can technically obscure a face, and they don't all hide identity the same way. Picking the wrong one for the job is the single most common reason a "blurred" face still reads as recognizable to someone who knows the person. **Mosaic Blur** replaces the covered area with flat, uniformly colored blocks. Per the manual, it's explicitly built for "hiding the face of anonymous witnesses," and that block-replacement approach is exactly why it's the safer default: once a region is fully broken into large enough blocks, there's genuinely no edge or gradient information left in that area for a viewer, or an algorithm, to reconstruct fine detail from. **Gaussian Blur** softens and smears pixels into their neighbors using a bell-curve falloff, which produces a gentler, more cinematic-looking blur. It's a fine tool for softening a background or hiding a logo where "not sharp" is good enough. For hiding a face specifically, Gaussian Blur at a moderate strength can leave more recoverable structure, jawline, hairline, general facial proportions, than an equivalently sized Mosaic Blur, because it's smoothing existing detail rather than replacing it outright. **Box Blur** works similarly to Gaussian but with a flatter, less naturalistic falloff, and it's rarely the first choice for either a cinematic soft look or a privacy blur. It sits between the other two in practical terms and is worth knowing exists more than it's worth reaching for on this specific task. | Filter | How it works | Best for | Weakness for hiding a face | | --- | --- | --- | --- | | Mosaic Blur | Replaces area with flat color blocks | Anonymizing a face or plate, Blackmagic's own recommended tool | Visually obvious, unmistakably "censored" look | | Gaussian Blur | Smooths pixels using a bell-curve falloff | Softening backgrounds, gentler stylistic blur | Can leave recoverable structure at moderate strength | | Box Blur | Flat-falloff smoothing | General softening | Rarely the best tool for either job | Here's the part worth taking seriously rather than skimming past: **neither filter is a perfect guarantee, and treating a light blur as bulletproof anonymization is a mistake worth knowing about before you rely on one.** A 2016 study out of Cornell Tech and Cornell University, titled ["Defeating Image Obfuscation with Deep Learning"](https://arxiv.org/abs/1609.00408) by Richard McPherson, Reza Shokri, and Vitaly Shmatikov, trained neural networks specifically against "mosaicing (also known as pixelation), blurring (as used by YouTube), and P3," a privacy photo-sharing system, and reported success at re-identifying faces and objects hidden behind all three. The researchers weren't attacking a hypothetical weak setting either; they tested obfuscation strengths similar to what platforms actually ship by default. That doesn't mean blurring is pointless. It means the strength of the blur matters far more than which filter you pick, and a heavily pixelated Mosaic Blur applied consistently across every frame the face appears in is meaningfully safer than a light touch you dialed in by eye until it "looked blurry enough." Push Pixel Frequency until you genuinely cannot tell who the person is from a single still frame pulled at random from the clip, not just while it's moving at 24 or 30 frames per second. **If your privacy stakes are real, whether it's a source protecting their identity, a minor, or someone who explicitly asked not to be shown, treat the blur as risk reduction, verify it by pulling a still frame and looking hard at it, and don't assume a filter default is strong enough just because it renders instantly.** ## How do you blur a face automatically with Magic Mask? Magic Mask is Resolve Studio's answer to a Power Window that just isn't precise enough, and for a face specifically, that means hair, ears, jawlines, and a head turning in profile, all the soft or irregular edges a simple circle or ellipse either misses or has to oversize to stay safe. Introduced in [DaVinci Resolve 17](https://www.premiumbeat.com/blog/resolve-17-magic-mask-tool/), Magic Mask uses the DaVinci Neural Engine to generate a mask from a single brush stroke rather than a hand-drawn shape. PremiumBeat's Lewis McGregor described the flexibility of the tool directly: **"With the Magic Mask, you can either generate a mask for the entire person or switch the operation to focus on a specific feature, from the arms to the person's shoes."** For this task, that means you can mask the whole head, or narrow it specifically to the face if you want hair to stay visible while identity stays hidden. Here's the workflow: 1. **Switch to the Color page** and select your clip. 2. **Open the Magic Mask panel** in the toolbar and choose the Add Click tool. 3. **Click directly on the face** you want to hide. The Neural Engine generates an initial mask around the head or face and begins tracking it automatically, frame by frame. 4. **Refine with Add and Subtract clicks** if the mask grabs too much background or misses part of the face on a difficult frame. 5. **Add a serial node**, drag Mosaic Blur onto it, and connect the Magic Mask node's alpha output into that node's key or mask input. 6. **Scrub the full clip** to confirm the mask holds through any head turns or brief occlusions before finalizing the blur strength. The tradeoff is processing cost. Magic Mask is a Neural Engine pass per clip, per face, and on longer footage or older hardware, that adds real time to your session. DaVinci Resolve 21, released June 2026, addressed part of that specific pain point directly. Per [CG Channel's coverage of the 21.0 release](https://www.cgchannel.com/2026/06/blackmagic-design-releases-davinci-resolve-21-0/), **"Magic Mask, DaVinci Resolve's AI-powered tool for automatically isolating people or objects in video footage, gets a new Render in Place option."** In practice, Render in Place bakes the tracked Magic Mask matte into a cached node, so once you're satisfied with the mask, playback and grading downstream of it no longer has to re-run the Neural Engine analysis on every frame. For a face blur specifically, that means you can lock the mask once, then adjust your Mosaic Blur's Pixel Frequency or tweak your grade elsewhere on the timeline without the whole pipeline bogging down every time you scrub. | Situation | Power Window | Magic Mask | | --- | --- | --- | | Static or gently moving face | Faster to set up, no GPU cost from Neural Engine | Overkill for this case | | Head turning through profile, hair moving | Requires manual keyframe correction on hard frames | Tracks the actual edge shape, less manual work | | Long clip, older or modest GPU | Lighter, real-time friendly | Slower first pass, faster after Render in Place | | Free version | Only option | Not available | **Magic Mask isn't a universal upgrade over a Power Window for this task, it's the better tool specifically when a face's actual edges, hair, jawline, ear, matter more than a fixed circle can approximate, and Render in Place in DaVinci Resolve 21 removes most of the performance penalty that used to make it painful on longer clips.** Reach for it when a Power Window keeps leaving a fringe of visible hair or jawline around your blur, not as your default starting point for every face in every shot. ## How do you blur a face on the Fusion page instead of the Color page? Everything above happens on the Color page, but Resolve gives you an entirely separate, fully free path through Fusion, and it's worth knowing even if you never touch Fusion for anything else, because it's the more flexible option when a face shares screen space with something complicated behind it. Per [Creative Video Tips' walkthrough of combining AI and manual masks in Resolve Fusion](https://creativevideotips.com/tutorials/combine-ai-manual-masks-in-resolve-fusion), the basic version of this workflow is genuinely simple: 1. **Open the Fusion page** and add a Blur node downstream of your MediaIn. 2. **Add an Ellipse mask** and position it over the face. 3. **Connect the Ellipse's mask output** into the Blur node's Effect Mask input, so the blur only applies inside that shape instead of the whole frame. 4. **Add a Tracker node**, place its tracking point on a stable, high-contrast feature of the face like the bridge of the nose or a corner of the eye, and track forward. 5. **Set the Tracker's operation to move the Ellipse's center**, so the mask follows the tracked point frame by frame instead of sitting in a fixed screen position. The advantage of doing this in Fusion rather than on the Color page is control. A Fusion Blur node's strength and the Ellipse mask's size, softness, and position are all independently keyframeable, and you can layer a second Ellipse with a Subtract operation if part of the masked area needs to stay sharp, something you'd have to build with a second Power Window and a serial node subtraction on the Color page. For a genuinely difficult shot, where the face turns fully to profile or gets crossed by another person repeatedly, you can combine Magic Mask's AI-generated matte with a manual Fusion mask rather than choosing one or the other. Route the Color page's Magic Mask output into Fusion as a matte, then use it to drive an Ellipse-refined Blur node, giving you the Neural Engine's tracking accuracy with Fusion's finer manual control over the mask's exact shape on the frames where the automatic result isn't quite tight enough. **The Fusion page's Blur-plus-Ellipse-plus-Tracker combination does the identical job as a Power Window and Mosaic Blur, and it's fully free in every version of Resolve, but it costs you more manual setup time in exchange for finer control over the mask shape and blend.** Reach for it when a Power Window's fixed circle or ellipse genuinely doesn't fit the shape of what you're hiding, not as your default starting point, since the Color page workflow gets you to the same result faster on a typical face. ## How do you blur more than one face in the same shot? A single interview subject is the easy case. A street scene, a classroom, or a group shot with several faces to hide is where a lot of tutorials quietly stop, because the answer isn't one bigger window, it's several independent ones running in parallel. The core rule: **one Power Window or one Magic Mask stroke tracks one face.** If two people move independently, one mask trying to cover both will either miss one of them the moment they separate, or stay oversized and blur far more of the frame than you need. Each additional face needs its own tracked window or its own Magic Mask click, and its own node in the chain if you want independent control over blur strength per face. Here's how that scales in practice: 1. **Add a serial node per face**, right-clicking your last node and choosing Add Node, Add Serial, once for each additional person. 2. **Draw or click a separate mask on each new node**, a Power Window for a simple case or a Magic Mask stroke for a harder one, and track each independently. 3. **Apply Mosaic Blur to each node individually.** You can share the same settings across all of them, or vary Pixel Frequency per face if some are closer to camera and need a coarser block size to fully obscure at that scale. 4. **Use Append Node to Selected Clips** if the same group of faces appears across multiple takes of the same locked-off shot, which copies your node structure and effect settings, though not the tracking data itself, onto each selected clip. For a genuinely crowded shot, five or more faces moving through frame at different times, manually tracking each one individually becomes a real time sink, and this is exactly the situation where a dedicated automatic tool starts to pull ahead of Resolve's native workflow, covered next. | Number of faces | Recommended approach | | --- | --- | | 1-2, mostly static or gently moving | Power Window plus Cloud Tracker per face | | 2-4, some rotation or occlusion | Magic Mask per face, Studio required | | 5+, moving independently through a busy scene | Consider a dedicated automatic face-detection plugin | **A crowded shot with five or more independently moving faces is the point where Resolve's manual tracking workflow, built one mask at a time, genuinely stops scaling, and that's worth admitting honestly rather than pretending a single Power Window can chase a crowd.** Know that threshold before you're two hours into hand-tracking a protest scene one face at a time. ## Can DaVinci Resolve blur faces automatically, with no tracking at all? Not natively, and it's worth being direct about that rather than implying otherwise. DaVinci Resolve 21's Neural Engine does a lot with faces already: facial recognition for organizing footage in the media pool, an AI Face Age Transformer that ages or de-ages a face, and an AI Face Reshaper for adjusting facial proportions. None of those is a "find every face in this clip and blur it automatically" button. Even Magic Mask, the closest native tool, still requires one click per face to start tracking. For genuinely automatic, no-click face detection and blurring across an entire clip, that's the specific job a third-party plugin was built to solve. **BCC+ Witness Protection ML**, part of Boris FX Continuum since the 2024.5 release, runs inside Resolve as an OpenFX effect. Per [Boris FX's own documentation](https://borisfx.com/documentation/continuum/bcc-witness-protection-ml/), the filter **"automatically finds human faces in every frame of the filtered source and generates an oval mask for every face that it finds."** It offers four independent ways to obscure identity, blur, mosaic, brightness and contrast reduction, and tint, and you can mix them for a custom look. On the Blackmagic user forum, a thread specifically about the best tool for face obscuring in Resolve 19 Studio lands on a similar split: Magic Mask works well for a handful of faces, IntelliTrack is the better choice for tracking through smaller obstacles, and Witness Protection is named directly as the option built for volume, machine-learning-driven detection across a whole crowd rather than one click per person. This is also worth naming plainly against the broader landscape of AI tools people reach for around DaVinci Resolve. Automatic face-blurring plugins like Witness Protection solve a narrow, specific job: find faces, hide them, move on. They don't teach you anything about the software itself. That's a genuinely different category from an assistant like TryUncle. **TryUncle is the on-screen assistant for DaVinci Resolve on macOS, ask in plain words and Uncle points at the exact control on your screen.** Where a face-blur plugin automates a task for you, Uncle watches your actual project and walks you through doing the task yourself, which matters the first few times you're setting up a Power Window track and want to understand what each checkbox is actually doing rather than just clicking through a preset. | Tool | What it does | Cost | Best for | | --- | --- | --- | --- | | Power Window + Cloud Tracker | Manual mask, manual track | Free | One or two faces, simple motion | | Magic Mask | AI-assisted mask from one click, manual track start | Studio license | A handful of faces, hair or rotation | | IntelliTrack | AI-assisted tracking of a manually placed point | Studio license | Faces with occlusion or busy backgrounds | | BCC+ Witness Protection ML | Fully automatic detection and masking, no clicks | Paid third-party plugin | Crowds, many faces, high volume work | **No feature inside DaVinci Resolve itself, free or Studio, finds and blurs every face in a shot with zero clicks, and knowing that up front saves you from hunting through menus for a button that doesn't exist.** If that's genuinely the job you need done at scale, a dedicated plugin like Witness Protection is the honest answer, not a hidden Resolve setting. ## Why does the blur flicker, pop off, or leave the face half-visible? Every symptom in this category traces back to the tracker, not the blur filter itself, and each has a specific, known fix. **The blur is solid, then suddenly gone for a few frames.** That's a break in the track, usually occlusion, a hand or another person crossing in front of the face, or a rapid head turn that momentarily loses the tracked features entirely. Switch from Clip mode to Frame mode in the Window Tracker palette, park on the last good frame before the gap, and hand-key the window's position through the blocked frames rather than trusting automatic tracking to solve a problem it has no data for. **The blur slides gradually off the face over several seconds.** That's drift, and it's almost always caused by too many transforms enabled at once, or tracking points sitting on smooth, low-contrast skin instead of a distinctive edge. Uncheck Zoom, Rotate, and Perspective 3D if the head isn't genuinely doing all three, and if it still drifts, nudge the tracking area toward a higher-contrast feature like a hairline, an eyebrow, or a collar seam nearby, then reanalyze. **The blur turns on or off at the wrong moment, or never turns off when you want it to.** This is a keyframe problem, not a tracking one. In a Creative COW forum thread specifically about this exact task, colorist Marc Wielage laid out the standard fix directly: **"use a power window to constrain the blur, then use Opacity to allow the blur or remove it as you see fit."** Fellow editor Glenn Sakatch walked through the mechanics in the same thread: open the keyframe editor, find the window's opacity attribute, and set a keyframe at 0 where the blur should be off and 100 where it should be on, letting Resolve interpolate the transition between them. A third editor in that thread, Michael Gissing, described an even simpler variant for a quick fade: ramp the blur's own strength value directly, from a baseline where nothing is blurred up to the value that hides the face. **The blur looks fine while scrubbing but reveals a gap or seam on export.** This is almost always a reduced-resolution preview problem. Set Timeline Playback Resolution to full under the Playback menu before you make a final judgment call on any tracked effect, since a track that looks locked at half or quarter preview resolution can reveal drift or a visible edge once every pixel is actually being rendered. | Symptom | Likely cause | Fix | | --- | --- | --- | | Blur vanishes for a few frames | Occlusion or a rapid head turn breaking the track | Frame mode, hand-key through the gap | | Blur slides gradually off the face | Too many transforms enabled, or low-contrast tracking points | Uncheck Zoom/Rotate/Perspective 3D; retrack off a higher-contrast feature | | Blur turns on/off at the wrong frame | Missing or misplaced opacity keyframes | Keyframe the window's Opacity or the blur's own strength value directly | | Blur looks solid in preview, reveals a gap on export | Reduced Timeline Playback Resolution | Set playback resolution to full before final review | | Track never locks on at all | Upstream noise reduction or lens correction altering pixels the tracker reads | Bypass upstream effects, track the clean plate, then reapply them | **A blur that flickers or pops off isn't a broken filter, it's an honest signal that the tracker underneath it lost the face, and the fix is almost always in the Tracker palette, not the Mosaic Blur settings.** If you're chasing a specific drift or occlusion pattern that doesn't match anything above, the full diagnostic tree in our guide to [Power Window tracking that keeps breaking](https://tryuncle.com/learn/davinci-resolve/davinci-resolve-power-window-keeps-losing-tracking) covers every remaining edge case, including anamorphic footage and upstream node ordering. And the same [keyframe editor and Clip mode versus Frame mode mechanics](https://tryuncle.com/learn/davinci-resolve/how-to-keyframe-effects-in-davinci-resolve) that fix a stubborn blur transition are the same ones you'll use for any hand-corrected animation in Resolve. If you're mid-session and can't tell whether what you're looking at is a drifting track, a missing keyframe, or an upstream node problem, that's exactly the kind of question Uncle can answer by looking at your actual project on screen instead of you re-reading a settings panel from scratch. ## Is a blurred face actually enough to hide someone's identity? This is worth answering honestly before you rely on a blur for anything with real stakes, because the answer isn't a simple yes. **A blur strong enough to remove all visible facial detail is meaningfully harder to reverse than a light one, but neither is an absolute guarantee, and you should know that going in.** The clearest evidence for this is a 2016 paper out of Cornell Tech and Cornell University, ["Defeating Image Obfuscation with Deep Learning,"](https://arxiv.org/abs/1609.00408) by Richard McPherson, Reza Shokri, and Vitaly Shmatikov. Its abstract states the finding directly: **"We demonstrate that modern image recognition methods based on artificial neural networks can recover hidden information from images protected by various forms of obfuscation."** The techniques they specifically targeted are "mosaicing (also known as pixelation), blurring (as used by YouTube), and P3," the exact same two filter families available inside Resolve FX Blur. That paper is a decade old at this point, and the neural networks capable of this kind of re-identification have only gotten more capable since 2016, not less. The practical takeaway isn't "don't bother blurring," it's "push the blur harder than feels necessary, and verify it on a still frame rather than trusting how it looks while playing back." **WITNESS**, a nonprofit that trains journalists and human rights activists on documenting events safely, has published guidance on exactly this problem for over a decade, starting with early coverage of [YouTube's built-in face-blurring feature](https://blog.witness.org/2017/08/introducing-youtubes-updated-blurring-feature/). Their broader guidance treats visual anonymization as a default precaution, not an afterthought, for anyone appearing in footage of a protest, a sensitive event, or anywhere being identified could put someone at real risk, and they note a genuinely important adjacent point: blurring a face in the footage you publish doesn't remove metadata, like timestamps and GPS location, still embedded in the original file if that original file is ever seized or subpoenaed. A blur protects the video you share. It does nothing for the source file sitting on your drive unless you handle that separately. Practical guidance that follows from all of this: 1. **Push Pixel Frequency on Mosaic Blur further than looks necessary.** If you can identify the person from a single paused frame, the blur isn't strong enough yet, regardless of how it reads in motion. 2. **Apply the blur consistently across every frame the face appears in.** A single unblurred frame, even for a fraction of a second at a hard cut or a tracking gap, defeats the entire purpose. 3. **Treat a blur as risk reduction for real safety stakes, not a guarantee.** For a source, a minor, or anyone whose exposure carries genuine consequence, combine a strong blur with the same file-security practices WITNESS recommends: strip metadata from the original, and don't assume the published, blurred version is the only copy that matters. **The honest answer to "is a blurred face untraceable" is no, not with certainty, and treating it as risk reduction rather than a guarantee is the more responsible way to think about this feature.** That doesn't make blurring pointless. It makes doing it properly, strong Pixel Frequency, full coverage, verified on a still frame, meaningfully more important than clicking the effect on and moving to the next task. ## Do you need to blur faces for YouTube, client work, or journalism? The honest answer is that it depends on where you and your subjects are, what the footage is for, and whether anyone in frame clearly consented to appear, and none of what follows is legal advice. **Under the EU's GDPR**, a person's image in video counts as personal data, and per [Gallio's overview of face blurring in digital media production](https://gallio.pro/blog/blurring-faces-and-license-plates-in-digital-media-and-video-production/), processing that data without a proper legal basis, meaning consent, a legitimate interest, or another Article 6 justification, can carry real regulatory consequences, and the same overview notes that similar redaction expectations show up in California's CPRA and in the UK's ICO guidance on filming in public. None of these frameworks require blurring every bystander in every shot; most carve out exceptions for things like journalism, incidental capture in public spaces, or legitimate documentary purposes. But "I filmed it in a public place" is not automatically a blanket exception everywhere, and the safest working assumption for anyone whose face is identifiable and who didn't sign a release is to either get clear consent on camera or blur them out. For journalism and activism specifically, the calculus shifts from "is this legally required" to "could this footage put someone at risk," and that's the exact problem WITNESS built their guidance around. Someone filming a protest on a phone and posting it within the hour isn't operating with an editor's review process behind them, and unredacted footage from exactly this kind of situation has been used in real, documented cases to identify and locate people who appear in it. That's not a hypothetical GDPR compliance question, it's a direct safety one, and it's the reason organizations that train journalists on this treat blurring bystanders and vulnerable subjects as a default, not a special case. For everyday client and YouTube work, here's a practical framework: | Situation | Blur recommended? | | --- | --- | | Interview subject who signed a release | No, unless they specifically requested it | | Bystander clearly visible and identifiable in the background | Yes, unless incidental and brief | | Minor appearing without a parent's explicit consent for that footage | Yes | | Someone who verbally asks you not to include them | Yes | | Protest, activism, or any footage where identification carries risk | Yes, treat as default | | Crowd shot where no individual is the visible subject | Usually not required, use judgment | **When someone didn't clearly consent to being identifiable in footage you're publishing, blurring their face is the lower-risk default, and reaching for it costs you a few minutes of tracking against a real, if hard to quantify, downside if you skip it.** This is one of the recurring questions we see inside our editing community from people cutting run-and-gun interview footage, corporate event coverage, or street content, and the answer is almost always the same: when you're unsure, blur it, and ask forgiveness from nobody because there's nothing to forgive. ## How do you make sure the blur survives your export? A blur that looks correct in the Color page viewer can still fail to show up correctly in your final render, and the reasons are worth checking before you ship anything with real privacy stakes riding on it. **Set Timeline Playback Resolution to full before your final review.** Resolve, by default on demanding timelines, renders a lower-resolution preview during scrubbing and playback to keep things responsive. A tracked blur that looks seamless at half or quarter resolution can reveal a gap, a seam, or a frame where the tracker briefly lost the face once you're actually looking at full-resolution pixels, and export always renders at full resolution regardless of what your preview showed you. **Check your Delivery page render settings match your review resolution.** If you graded and reviewed at a lower project or timeline resolution than your final delivery format, a proportionally sized Power Window can still track correctly in relative terms, but always re-scrub the final export itself, not just the timeline, before sending it anywhere. **Confirm the node with your blur is upstream of any output LUT or sharpening you've added.** A sharpening filter applied after Mosaic Blur can, in rare cases, slightly recover edge detail at the boundary of your blocks. It's a minor effect compared to the blur strength itself, but it's a real reason to keep your blur node's position in the node chain deliberate rather than accidental. **If you're delivering to multiple formats or platforms**, verify the blur in each one separately rather than assuming a pass on one export guarantees the rest. A render for YouTube at one resolution and a vertical crop for a different platform can reframe the shot enough that a Power Window sized for the original composition no longer fully covers the face in the cropped version. This matters even more if you're pulling frames out for a thumbnail or a social clip separately from the main export, since a thumbnail generator can sometimes grab a frame from a section of the timeline you didn't expect. If you're exporting specifically for YouTube, our [DaVinci Resolve export settings for YouTube guide](https://tryuncle.com/learn/davinci-resolve/davinci-resolve-export-settings-youtube) covers the format and resolution settings that determine exactly what gets baked into that final file. **Always play back the entire finished export file itself, not just the timeline inside Resolve, before you publish or send it anywhere the blur actually matters.** A render that completes without an error message isn't proof the blur is intact for every single frame it needs to cover. It's proof the file encoded successfully. Those are different claims, and confusing them is how an unblurred frame ends up in a published video. ## Free version vs Studio: which face-blurring tools do you actually have? Every core building block for blurring a face, the Power Window, the Cloud Tracker, and every Resolve FX Blur filter including Mosaic Blur and Gaussian Blur, ships in the free version of DaVinci Resolve with no restriction. What Studio adds is speed and convenience on harder shots, not a capability the free version lacks entirely. | Tool | Free version | Studio | | --- | --- | --- | | Power Window (circle, ellipse, polygon, curve) | Yes | Yes | | Cloud Tracker | Yes | Yes | | Mosaic Blur, Gaussian Blur, Box Blur | Yes | Yes | | Keyframe editor for opacity/blur ramping | Yes | Yes | | Fusion Blur node, Ellipse mask, Tracker node | Yes | Yes | | Magic Mask | No | Yes | | IntelliTrack | No | Yes | | Render in Place (Magic Mask caching) | No | Yes | | BCC+ Witness Protection ML (third-party) | Requires separate Boris FX purchase | Requires separate Boris FX purchase | Studio's Neural Engine features, Magic Mask and IntelliTrack specifically, exist to save you manual correction time on the hardest cases: a face that turns fully to profile, gets briefly blocked, or needs a mask that hugs hair and jawline more precisely than a fixed geometric shape can. If your footage is mostly static interview setups, locked-off shots, or faces that move but don't rotate much, the free version genuinely covers the whole job with no compromise in the final result, just more of your own manual tracking and keyframe time on the harder shots. **Studio doesn't unlock a face-blurring capability the free version lacks, it trades your manual tracking and correction time for GPU processing time on the hardest shots, and that's a real tradeoff worth weighing honestly rather than assuming Studio is required for this task.** If you're doing this occasionally, on mostly simple footage, stick with the free version's Power Window and Cloud Tracker. If you're doing it regularly on footage with a lot of head movement, crowds, or occlusion, Studio's Magic Mask and IntelliTrack will save you real hours over a project. ## What does a full worked example look like, start to finish? Here's the whole process applied to four realistic shots, so you can see how the pieces above connect in the order they actually get used. **A locked-off interview, subject sitting still, no release signed.** You check motion first: no movement at all, camera and subject both static. You draw a circular Power Window over the face, add a serial node, drag Mosaic Blur onto it, and raise Pixel Frequency until the face is unrecognizable in a paused frame. No tracking needed. Total time: about ninety seconds. **The same interview, but the subject leans forward and turns their head to gesture at something off-camera around the six-second mark.** Now there's real motion and rotation. You draw the window on the clearest frame, which happens to be a few seconds in, open the Tracker palette, leave Pan and Tilt checked, and uncheck Zoom, Rotate, and Perspective 3D to start. You track forward and back from your anchor frame. The track holds cleanly through the lean, but drifts slightly during the head turn, since a plain Cloud Tracker has to solve for real rotation there. You have Studio, so you switch to IntelliTrack for that section instead, retrack, and it holds through the full turn without drift. **A street scene with three pedestrians crossing the frame at different times, no release from any of them, shot handheld.** This is the multiple-faces case with real camera motion layered on top. You add three separate serial nodes, one per person, and click each face individually with Magic Mask rather than fighting three independent Power Window tracks against a moving handheld background. Two of the three track cleanly start to finish. The third briefly crosses behind a parked car, and Magic Mask's tracking momentarily loses the face; you hand-correct that section in Frame mode across the four or five frames where the occlusion happens, then let automatic tracking resume once the person clears the obstruction. You apply Mosaic Blur to each of the three nodes, render in place to keep grading responsive, and export. **Archival protest footage you're editing for a documentary, several dozen faces total across a wide crowd shot, safety genuinely at stake for the people shown.** This is past the point where manual, one-face-at-a-time tracking makes sense. You reach for a dedicated automatic detection plugin like BCC+ Witness Protection ML, let it run its face-detection pass across the whole clip, review its output frame by frame since automatic detection on a crowd this size will miss a face here and there, and manually add a Power Window with Mosaic Blur for any face the automatic pass didn't catch. Before you export, you also strip location and timestamp metadata from your project's source files, since a blurred face in the published cut doesn't protect anyone if the original footage's metadata is exposed elsewhere. Four different starting points, all under the umbrella of "blur a face," and four genuinely different paths through the same underlying toolset, because the actual variable driving every decision was motion, headcount, and how much the stakes justified extra verification time, not anything about the subject matter itself. ## Quick troubleshooting reference Bookmark this table. It condenses every fix from this guide into one lookup. | Symptom | Likely cause | Fix | | --- | --- | --- | | Blur is fine, then vanishes for several frames | Occlusion or a fast head turn breaking the track | Frame mode, hand-key through the gap | | Blur slides gradually off the face | Too many transforms enabled, or low-contrast tracking points | Uncheck Zoom/Rotate/Perspective 3D; retrack off a higher-contrast feature | | Blur turns on or off at the wrong frame | Missing opacity or blur-strength keyframes | Keyframe the window's Opacity, or ramp the blur value directly | | Blur looks solid in preview, gap shows on export | Reviewing at reduced Timeline Playback Resolution | Set playback resolution to full before final judgment | | Track never locks on at all | Upstream noise reduction or lens correction altering the tracked pixels | Bypass upstream effects, track the clean plate, reapply after | | Mask misses hair, ears, or jawline | Power Window's fixed circle doesn't match the head's actual shape | Switch to Magic Mask (Studio) for a shape that tracks the real edge | | Second person accidentally gets blurred too | One oversized window covering two independently moving faces | Split into separate serial nodes, one tracked window per face | | Grading feels sluggish after adding Magic Mask | Neural Engine reprocessing the mask on every scrub | Use Render in Place (DaVinci Resolve 21) to cache the matte | | Blur strength looks fine in motion but face is identifiable on a paused frame | Pixel Frequency set too low for genuine anonymization | Raise Pixel Frequency until no feature survives on a still frame | | Magic Mask or IntelliTrack option is grayed out | Free version, not Studio | Confirm you're running DaVinci Resolve Studio | ## The verdict Blurring a face in DaVinci Resolve isn't one button, it's a Power Window, a tracker, and a blur filter, three tools that were all free the entire time, working together. Studio's Magic Mask and IntelliTrack make the hard cases, rotation, occlusion, crowds, faster to solve, but they're a convenience on top of a workflow the free version already handles completely for the majority of shots you'll actually run into. **Diagnose motion and headcount before you draw a single window, and this whole task turns into a short lookup instead of a guessing game.** That one habit, checking whether the face moves, turns, or gets occluded, and how many faces you're actually dealing with, decides your entire toolset before you've touched a single effect. Get the diagnosis right and Mosaic Blur, applied strong enough that a paused frame gives up nothing, does the rest. If you're mid-project and can't tell whether your track is drifting, whether Magic Mask is worth the setup time for this specific shot, or which node your blur actually needs to sit on, that's exactly the gap [TryUncle](https://tryuncle.com/?utm_source=learn&utm_medium=blog&utm_campaign=how-to-blur-a-face-in-davinci-resolve) is built to close, an assistant that looks at your real project on screen instead of a generic tutorial clip filmed on someone else's footage. And once your blur is tracking cleanly, our guide to [DaVinci Resolve's Power Window tracker losing tracking](https://tryuncle.com/learn/davinci-resolve/davinci-resolve-power-window-keeps-losing-tracking) covers the same drift and occlusion problems in more depth, for anyone building the tracked mask this entire workflow depends on. ## FAQ ### Is blurring a face in DaVinci Resolve free, or do you need Studio? Free. Power Windows, the Cloud Tracker, and every Resolve FX Blur filter, including Mosaic Blur and Gaussian Blur, work in the free version with no restrictions. Studio adds Magic Mask and IntelliTrack, both Neural Engine features, which make tracking and masking a moving face faster and more hands-off, but they are conveniences, not requirements. A free-version user with a Power Window, a tracker, and a blur node can blur any face a Studio user can. ### Why does my face blur pop off or flicker when the person turns their head? That's almost always a tracking problem, not a blur setting. A Power Window's Cloud Tracker follows contrast, and a turning head presents the tracker with a different set of pixels every few frames, which is exactly the kind of motion that causes drift or a hard break. Turn off Zoom, Rotate, and Perspective 3D if the head isn't genuinely rotating in 3D space relative to camera, and if it snaps off at one specific frame rather than sliding gradually, that's occlusion or a profile turn losing the tracked features entirely, which needs Frame mode hand correction rather than a checkbox fix. ### Can DaVinci Resolve blur faces automatically without you tracking anything? Not on its own. DaVinci Resolve 21's Neural Engine includes facial recognition for organizing media and AI tools like Face Age Transformer and Face Reshaper for retouching, but there's no built-in one-click 'find every face and blur it' function. The closest native workflow is Magic Mask, which still needs you to click on each face once. For genuinely automatic, no-click detection across every frame, that's a job for a third-party plugin like Boris FX's BCC+ Witness Protection ML, which runs inside Resolve as an OpenFX effect. ### Is a blurred face in a video actually untraceable? Not with certainty, and you should know that before you rely on it. A 2016 Cornell/Stanford study titled 'Defeating Image Obfuscation with Deep Learning' trained neural networks to re-identify faces hidden behind mosaic pixelation and blur, the same two techniques Resolve's Mosaic Blur and Gaussian Blur filters use. A blur strong enough that no detail survives, applied consistently across every frame, is far safer than a light one, but treat it as risk reduction, not a guarantee, especially for anything with real safety stakes. ### What's the actual difference between Mosaic Blur and Gaussian Blur for hiding a face? Mosaic Blur replaces the area with flat colored blocks and is Blackmagic's own suggested tool for this job, described in its manual as 'a simple, pixelated blur suitable for hiding the face of anonymous witnesses.' Gaussian Blur smooths and softens instead of blocking, which looks gentler but can leave more recoverable detail at a given strength. For anonymity, push Mosaic Blur's Pixel Frequency until no facial feature is guessable, not until it merely looks blurry. ### Do I legally need to blur faces in my YouTube or client videos? It depends on where you and your subjects are and what the footage is used for, and this isn't legal advice. Under GDPR, a person's face in video is personal data, and using it without a legal basis can carry real penalties. Organizations like WITNESS, which trains journalists and activists on protecting subjects on camera, treat blurring as a default precaution for anyone who didn't clearly consent to appear, especially at protests or sensitive events. When in doubt, get consent on camera or blur the face. ### Can I blur a face using only the Fusion page, without touching the Color page at all? Yes. Add a Blur node after your MediaIn, add an Ellipse mask connected to the Blur node's effect mask input, and attach a Tracker node to carry the ellipse across frames as the face moves. It's more manual setup than the Color page's Power Window and Mosaic Blur combination, but it's fully free, works identically in every version of Resolve, and gives you finer control over exactly which pixels get sampled into the blur. ## Sources - [Blackmagic Design DaVinci Resolve Manual: Mosaic Blur](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part3161.htm) - [Blackmagic Design DaVinci Resolve Manual: Resolve FX Overview](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part3150.htm) - [Blackmagic Design DaVinci Resolve Manual: Motion Tracking Windows](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part2864.htm) - [Blackmagic Design DaVinci Resolve Manual: Controls in the Window Tracker Palette](https://www.steakunderwater.com/VFXPedia/__man/Resolve18-6/DaVinciResolve18_Manual_files/part2873.htm) - [Boris FX: How to Blur Part of a Video in DaVinci Resolve, by Marco Sebastiano Alessi](https://borisfx.com/blog/how-to-blur-part-of-a-video-in-davinci-resolve-in/) - [Boris FX Documentation: BCC+ Witness Protection ML](https://borisfx.com/documentation/continuum/bcc-witness-protection-ml/) - [CG Channel: Boris FX Releases Continuum 2024.5](https://www.cgchannel.com/2024/06/boris-fx-releases-continuum-2024/) - [RedShark News: In-Depth With DaVinci Resolve 19's IntelliTrack and ColorSlice Features, by Blake Jones](https://www.redsharknews.com/in-depth-with-davinci-resolve-19s-intellitrack-and-colorslice-features) - [Creative COW Forum: track & blur a face for part of a shot (Marc Wielage, Glenn Sakatch, Michael Gissing)](https://creativecow.net/forums/thread/track-blur-a-face-for-part-of-a-shot/) - [Creative Video Tips: Combine A.I. + Manual Masks in Resolve Fusion](https://creativevideotips.com/tutorials/combine-ai-manual-masks-in-resolve-fusion) - [WITNESS Blog: Introducing YouTube's Updated Blurring Feature](https://blog.witness.org/2017/08/introducing-youtubes-updated-blurring-feature/) - [arXiv: Defeating Image Obfuscation with Deep Learning (McPherson, Shokri, Shmatikov, 2016)](https://arxiv.org/abs/1609.00408) - [Gallio PRO: Blurring Faces and License Plates in Digital Media and Video Production, by Mateusz Zimoch](https://gallio.pro/blog/blurring-faces-and-license-plates-in-digital-media-and-video-production/) - [PremiumBeat: A Focus on Resolve 17's New Masking Tool: Magic Mask, by Lewis McGregor](https://www.premiumbeat.com/blog/resolve-17-magic-mask-tool/) - [CG Channel: Blackmagic Design Releases DaVinci Resolve 21.0](https://www.cgchannel.com/2026/06/blackmagic-design-releases-davinci-resolve-21-0/) - [Cutsio: How to Use Magic Mask Render in Place in DaVinci Resolve 21](https://cutsio.com/blog/davinci-resolve-21-magic-mask-render-in-place) - [Blackmagic Forum: Best Tool for Face Obscuring in DaVinci Resolve 19 Studio](https://forum.blackmagicdesign.com/viewtopic.php?f=21&t=205252)