RESOLVE FX · STUDIO AI LENS
AI CineFocus.
Refocus the scene.
Build the lens.
From an AI depth map to a believable focus pull. Understand the four CineFocus views, separate depth quality from bokeh quality, repair difficult edges and keep grain consistent through defocus.
Internal / Alpha
Distance + DoF
Aperture + Grain
Switch the lab to Depth, click the subject, then widen the aperture. The model explains relationships without pretending to be Resolve’s renderer.
Manual focus: 0.52. Click the scene or choose a target.
Lens character in the model
These controls illustrate relationships only; their scale is not Resolve’s scale.
Bokeh Quality, Depth Quality, aberrations, edge reconstruction and grain processing are explained below but deliberately not faked by this browser model.
Teaching model, not Resolve: the scene, depth values and blur math are synthetic. Canvas blur and focus peaking only approximate depth-aware editing; the scene hit-test uses synthetic shapes, not image analysis. This model does not reproduce Resolve’s AI depth, Bokeh Quality, lens physics, grain tools or parameter scales.
DaVinci Resolve Studio 21 · depth-aware virtual lens
4 views · 9 toolsets01THE DEPTH-OF-FIELD MODEL
Three stages must agree: depth, focus and lens.
CineFocus is useful because it does not treat the frame as one flat blur surface. The teaching lab illustrates the relationships, while the official interface and guide remain the authority for Resolve’s actual rendering.
Depth comes first, then the lens.
CineFocus does not apply one uniform blur to the whole frame. With Internal selected it estimates depth from a normal 2D image, assigning scene regions to different distances. Focus Distance and Depth of Field then define the depth band that stays sharp, while the virtual aperture shapes defocus outside that band.
Focus Distance is position; Depth of Field is width.
These controls are not interchangeable. Focus Distance moves the focal plane through the depth map. Depth of Field widens or narrows the range of depths considered sharp. A wider DoF therefore is not the same as a different Focus Distance: the focus center stays put while the accepted range around it changes.
Bokeh is virtual-lens behavior, not just blur radius.
Shaped Bokeh, blade count, Curvature, Pinwheel and Anamorphic Lens change the geometry of out-of-focus highlights. Lens Properties and Lens Distortion add imperfections associated with real optics. This is why CineFocus is useful when believable depth and defocus character matter, rather than merely reducing detail.
Official CineFocus model and control descriptions: DaVinci Resolve 21 New Features Guide · pp. 45–50 ↗
02THE FOUR VIEW MODES
Do not tune everything in Result mode.
Blackmagic organizes CineFocus from left to right: Depth → Focus → Aperture → Result. Each view answers a different diagnostic question.
Verify what the AI thinks the scene depth is.
Depth Mode displays the depth map. In the official workflow, brighter regions are closer to camera, mid-gray regions sit in the mid-field and darker regions are farther away. Check this before polishing bokeh: if the depth model is wrong, the defocus will be wrong too.
Choose the focal plane or focus range.
In Focus Mode the in-focus region remains colored while out-of-focus areas are subdued. The OpenFX overlay lets you click the scene to focus and drag the split-prism to select a focus range. This is more direct than judging Focus Distance by the number alone.
Inspect the virtual-lens model.
Aperture Mode visualizes the light path around the focal plane and the bokeh size at different distances. It is especially useful for Aperture Size, aperture shape and anamorphic character—settings that can look restrained in one shot and excessive in another.
Return to the final image and judge it in motion.
Result shows the finished CineFocus image while controls remain available for adjustment. Always return here after the technical modes: a correct depth map does not automatically mean a natural result. Check edges, highlights, motion, grain and a short export at delivery resolution.
The mode descriptions above follow Blackmagic’s official Resolve 21 CineFocus workflow. The site lab echoes these modes with synthetic graphics only; it does not reproduce the OpenFX overlay or Resolve AI model.
03THE ACTUAL CONTROLS
Treat each toolset as a separate stage.
Eight local screenshots plus the official guide were used to map the visible controls. Dimmed child rows are documented because they become available when their parent switches are enabled.
01 · Focus
Focus, sharp-depth range and defocus character.
- Bokeh Quality
Rendering mode for bokeh. Enhanced prioritizes detail and can be slower at large apertures; Better is softer and more efficient; Faster applies stronger optimization. It is not the Depth Map quality setting.
CheckUse Faster or Better while tuning, then compare Enhanced on a short section before final render. Do not judge the mode from a single still only.
- Aperture Size
Changes the virtual aperture: larger values increase defocus outside the sharp range and enlarge out-of-focus highlights. It is a primary creative control, but it does not replace Depth of Field.
CheckEstablish Focus Distance and DoF first, then increase Aperture Size. Very large apertures quickly expose depth-boundary errors and artificial-looking bokeh.
- Focus Distance
Sets the focal plane from Depth Map data. It is a percentage inside the depth model, not physical meters or feet. It can be animated to create a focus pull.
CheckFor a static subject it is often easier to click the desired region using the OpenFX overlay. For a controlled rack focus, place keyframes deliberately and inspect the transition between them.
- Depth of Field
Defines the depth range considered in focus. CineFocus can widen it artificially so a subject remains sharp without globally reducing near/far defocus.
CheckIf hair, shoulders or a physically deep subject partially fall out of focus, try adjusting DoF before immediately reducing Aperture Size.
- Track Focus to Point
Makes focus follow a point selected in the FX Overlay. The FX Tracker can follow that point throughout the clip.
CheckCheck the track through occlusions and when the subject approaches frame edges. Point tracking does not repair an incorrect depth map around the subject.
- No Foreground Blur
Removes blur in front of the focal plane, keeping the foreground sharp. The official guide notes that this can make it easier to keep a foreground subject sharp and can improve performance.
CheckUse it only when it matches the shot. If the foreground should naturally fall out of focus, enabling it will undermine the optical look.
- Focus Peaking
Enables focus-peaking lines that highlight regions considered sharp. It is a diagnostic view, not part of the final image.
CheckUse it to verify the chosen range, then turn it off and judge the actual Result. Peaking does not replace viewing at 100%.
- Boost Highlights
Raises highlight brightness as regions mix through defocus and can make bokeh stand out against dark backgrounds. The result also depends on the project color-management setup.
CheckWatch for clipping and do not use Boost Highlights as a general exposure correction. Compare scopes and the final render.
02 · Depth Map
Depth source and map repair.
- Source · Internal / From Alpha Input
Internal builds the Depth Map inside CineFocus. From Alpha Input uses a depth map prepared elsewhere and supplied through the alpha input, allowing a controlled depth pass to replace the internal AI estimate.
CheckDo not switch to an external source simply because it exists. It is useful when you can genuinely prepare a more accurate map and route it into the node reliably.
- Quality · Faster / Better
Quality setting for the internal depth analysis specifically. Faster prioritizes speed; Better prioritizes depth-model and boundary quality. It is separate from Bokeh Quality.
CheckIf hair, hands or thin foreground boundaries are unstable in Faster, compare Better before manually repairing the map.
- Invert
Inverts the Depth Map, swapping the near/far interpretation. Use it when an external depth pass has opposite polarity or the map is semantically reversed.
CheckDo not fix a local map error with a global Invert. If one object is wrong while the rest of the scene is correct, use a more targeted correction.
- Adjust Map Levels · Far Limit / Near Limit / Gamma
Enables manual remapping of the Depth Map. Far Limit adjusts the far/black end, Near Limit the near/white end, and Gamma redistributes intermediate depths.
CheckAdjust levels while viewing Depth Mode and make sure separate depth planes do not collapse into one tone. An overly aggressive map reduces the smoothness of focus transitions.
- Post-Processing · Post Filter / Contract / Expand / Blur
Post-processing can smooth the map along image contours, contract or expand foreground regions and soften jagged boundaries. Depth-map Blur is a repair tool and can itself create artifacts around object edges.
CheckUse the minimum correction necessary. If the depth map needs heavy blur, revisit the source analysis, Better quality and the separation between depth planes.
- Separate Layers
Strengthens separation between nearby depth planes when one object sits in front of another. The official guide specifically says to keep this at 0.0 when using an external Depth Map.
CheckIncrease it only until the planes separate clearly enough. Too much can turn continuous depth into visible steps.
03 · Lens Properties
Imperfections of real optics.
- Chromatic Aberration · Strength / Phase
Simulates different spectral components focusing at different depths. Strength sets the amount of color separation; Phase changes the relative color spread according to the simulated optics.
CheckAdd it only after depth and basic defocus are working. Too much quickly reads as a digital RGB split rather than lens character.
- Spherical Aberration · Lens Model / Strength
Simulates how a large aperture can spread focus over a range of depths. Lens Model selects the modeled construction and Strength sets the amount of dispersion.
CheckDo not use it to fix a wrong Focus Distance. It is a lens-character control, not a focal-plane selector.
04 · Aperture
Bokeh shape and anamorphic character.
- Shaped Bokeh · Number of Blades / Rotation / Curvature / Pinwheel
Enables aperture-shape control. Number of Blades changes blade count, Rotation turns the shape, Curvature bends blade edges and Pinwheel exposes the blade-tip character that can create a pinwheel-like bokeh shape.
CheckJudge it on real point highlights and do not make the shape an end in itself. In a scene without visible highlights, differences may be subtle.
- Anamorphic Lens · Anamorphism
Enables an anamorphic bokeh shape. Anamorphism sets the amount of stretch and changes the aspect ratio of out-of-focus highlights.
CheckMatch it to the rest of the image character. Isolated stretched bokeh without compatible optical cues can look like an effect rather than a lens.
05 · Lens Distortion
Distortion toward the frame edges.
- Curved Lens · Curvature / Lateral Aberration / Phase
Adds artificial lens curvature. Curvature sets the geometric edge distortion, Lateral Aberration adds color distortion to blur near the frame edges, and Phase changes the relative color spread.
CheckInspect straight lines and faces near frame edges. This section stylizes the virtual lens; it is not a real-lens geometry correction tool.
06 · Foreground Edge Overlap
Depth boundaries are where synthetic DoF is easiest to spot.
- Overlap Preview
Shows where the blurred foreground overlaps the rest of the scene. It is a diagnostic view rather than the final image.
CheckUse it on difficult hair, railings, fingers and thin objects, where overlap artifacts are most visible.
- Overlap Boost
Makes overlapping foreground denser to help hide transition edges. It only has an effect where the foreground actually overlaps the background.
CheckIf the boundary becomes heavy or produces a dark/bright rim, reduce the value and compare Automatic Edge Handling.
- Automatic Edge Handling
Automatically creates more natural defocused overlaps at the cost of some background detail in those regions. Enabling it disables Overlap Boost.
CheckCompare both the subject edge and the background immediately behind it. Automatic cleanup can hide more background detail than you are willing to lose.
07 · Remove and Replace Grain
Do not let grain reveal the synthetic blur.
- Remove Grain · Grain Detection / Estimate Best Thresholds
Analyzes and removes source grain first so defocus is built from an image without partially blurred noise texture. Grain Detection offers analysis methods; AI Ultra Noise Reduction can estimate thresholds automatically.
CheckDo not treat this section as a general replacement for a dedicated NR workflow. Its purpose here is to keep texture consistent with the virtual depth of field.
- Split Luma Chroma Thresholds · Threshold · Preview Grain
Lets you separate luma and chroma grain thresholds, set the boundary between grain and image detail, and preview the detected texture.
CheckPreview Grain should not contain large amounts of real facial, hair, fabric or architectural detail. Otherwise you are removing structure, not just grain.
- Replace Grain After Defocus · Grain Replacement
Restores the removed grain after depth defocus has been calculated and controls how much texture is added back. This places grain over the processed image instead of blurring it together with the background.
CheckCompare a 100% view across sharp and defocused regions. Grain should read as one consistent layer rather than disappearing in the background or doubling up.
08 · Additional Grain
A new texture layer over the final image.
- Add Grain · Strength / Size / Softness / Saturation
Adds a separate grain layer over the entire image. Strength sets amount, Size the particle scale, Softness the grain sharpness, and Saturation its colorfulness. It is separate from restoring the source grain.
CheckIf Replace Grain After Defocus is already active, Additional Grain can double the texture. Decide whether you want the source grain restored, a new stylized layer, or both intentionally.
09 · Global Blend
Final result mixing comes last.
- Blend
Mixes the finished CineFocus output with the effect input. Lowering Blend restores original sharpness and texture over the already-calculated depth defocus; it does not physically reduce aperture or widen DoF.
CheckFirst get depth, edges, bokeh and grain into a coherent state at full effect. Use Blend only afterward as a creative mix.
04FROM IMAGE TO DEPTH-AWARE LENS
Give CineFocus a clean place in the node tree.
Keep base balancing separate. With Internal selected, CineFocus builds depth itself. With From Alpha Input, feed a prepared depth map and validate polarity before touching aperture styling.
The depth-aware lens node
Keep Source on Internal. Validate Depth Mode first, then set Focus Distance / DoF, aperture and edge handling. Optical styling and grain come after a reliable depth model.
With Internal selected, CineFocus builds its own depth model. Keep depth validation, optical styling and primary grading as separate mental steps even when they live inside one OFX panel.
Related Color workflow: node connections ↗ · AI Depth Map ↗
05BUILD IT IN RESOLVE
Four exercises that expose different failure modes.
These are practical workflows, not presets. Start from a reliable depth map, judge the shot in motion, and use a short delivery-resolution render as the acceptance check.
Turn an all-sharp image into depth-of-field footage
The primary official use case: strengthen subject separation without manually building multiple blur layers.
PrepareUse a clip with a readable foreground, subject and background. Start with Internal Depth Map, Bokeh Quality Better, and the Result/Depth/Focus modes.
- 01
In Depth Mode verify that the subject separates from the background and the major planes are not reversed.
- 02
Click the subject in Focus Mode or set Focus Distance, then widen Depth of Field only enough to keep the whole subject sharp.
- 03
Increase Aperture Size to visible but believable defocus; change bokeh shape only if the shot contains suitable highlights.
- 04
Return to Result and inspect hair, fingers, fine edges and the moving background; compare Better depth quality and Edge Handling if needed.
The subject stays sharp across its depth, the background falls out of focus by distance, and the boundary does not look like a cutout in a still or in motion.
Build a controlled focus pull between two planes
Animate Focus Distance so attention moves from foreground to background or the reverse.
PrepareChoose a shot with two clearly separated depth subjects. Make sure the Depth Map does not jump between them.
- 01
Set the first Focus Distance position and add a keyframe.
- 02
At the second moment move Focus Distance to the other subject and add another keyframe.
- 03
Adjust Depth of Field so the transition reads clearly while each intended subject can still sit inside the sharp zone.
- 04
Review the transition in real time or through cache/render; adjust keyframe timing when needed rather than only changing aperture.
Focus moves through depth smoothly without a depth-model jump or an accidental moment where both important subjects are equally soft.
Repair a difficult foreground boundary
Reduce a halo around hair, a thin object or foreground sitting in front of a strongly defocused background.
PrepareOpen the problem frame at 100% and switch between Depth Mode, Result and Overlap Preview.
- 01
First check the Depth Map itself: Better quality, Adjust Map Levels and minimal Post-Processing.
- 02
If two depth planes sit close together, cautiously increase Separate Layers. Keep it at 0.0 for an external Depth Map.
- 03
In Foreground Edge Overlap compare Overlap Boost with Automatic Edge Handling; do not treat them as two independent boosts.
- 04
Inspect several neighboring frames; fixing one still should not create boundary flicker in motion.
The boundary preserves the subject shape, background does not leak into the foreground, and no heavy synthetic rim appears.
Keep texture natural after strong defocus
Avoid a perfectly smooth background while the sharp subject still carries source noise or film grain.
PrepareUse source footage with visible texture at 100%. Set up depth and defocus before adding Additional Grain.
- 01
Enable Remove Grain and tune detection/threshold so Preview Grain contains texture without swallowing useful fine detail.
- 02
Tune CineFocus on the cleaned image.
- 03
Enable Replace Grain After Defocus and tune Grain Replacement.
- 04
Only if a separate creative layer is needed, add Additional Grain and verify that texture has not been doubled.
Grain reads as one layer over both sharp and defocused areas rather than as scene detail that strangely disappears with depth.
06CHOOSE THE RIGHT BLUR MODEL
CineFocus is powerful precisely because it is not a generic blur.
Use a simpler FX when you do not need scene depth. Fewer moving parts often produce a more predictable result.
| FX | What shapes the result | Use it when |
|---|---|---|
| AI CineFocus | Depth-aware virtual lens: AI Depth Map, focal plane, DoF, aperture/bokeh, edges and grain. | When different scene planes need different defocus and a focus pull may be required. |
| Box Blur ↗ | Uniform averaging without depth reconstruction; Strength, H/V, iterations, borders and alpha. | General softening, a local mask or technical blur when depth-aware optics are unnecessary. |
| Gaussian Blur ↗ | Smooth spatial blur over the whole signal or a selected region. | Neutral softness without bokeh, lens character or an AI Depth Map. |
| Lens Blur ↗ | Optical blur/bokeh shape without built-in scene-depth reconstruction. | When aperture shape and bokeh character matter but depth is supplied another way or is unnecessary. |
| Directional Blur ↗ | Angle-driven smear rather than a depth-of-field model. | Motion stylization, directional streaks and elongated blur. |
| Mosaic Blur ↗ | Discrete cells/pixelation instead of optical defocus. | Graphic concealment and a pixelated look; not a depth-of-field substitute. |
07DIAGNOSE THE STAGE, NOT THE SLIDER
Most failures belong to depth, focus, edges or texture.
01A bright or dark halo appears around hair or thin foreground.+
Check Depth Mode and Better quality first, then Post-Processing and Separate Layers. After that use Overlap Preview and compare Overlap Boost with Automatic Edge Handling. Do not fix edge geometry merely by reducing Global Blend.
02CineFocus focuses on the background instead of the subject.+
Inspect the Depth Map and near/far polarity. If the whole map is reversed, use Invert. If only one object is wrong, a global Invert will not help; repair the map or the external depth pass.
03Part of the subject is soft even though Focus Distance is correct.+
Increase Depth of Field to cover the subject’s actual depth. If the problem is only at the boundary, return to the Depth Map; if the whole subject is soft, check Focus Distance and Aperture Size.
04Bokeh looks like digital blobs or an aggressive filter.+
Reduce Aperture Size and Boost Highlights, temporarily disable Shaped Bokeh/aberrations, and establish a natural base defocus. Then add optical traits back one at a time.
05The focus pull jitters or jumps between depths.+
Separate two problems: Focus Distance animation and an unstable Depth Map. Check Depth Mode through the whole clip first, then inspect keyframe timing or Track Focus to Point.
06Grain disappears in the background but remains on the subject.+
Use Remove and Replace Grain: remove texture before defocus and restore it afterward. If Additional Grain is enabled, make sure it is not doubling the restored source grain.
07The effect is too heavy for interactive tuning.+
Use Bokeh Quality Faster/Better instead of Enhanced, Depth Quality Faster instead of Better during rough setup, and temporarily disable heavy grain repair. Restore final-quality choices before the validation render.
08An external depth map produces stepped or overly hard separation.+
Check the external pass range and polarity, then Far/Near/Gamma and minimal Post-Processing. Keep Separate Layers at 0.0 as the official guide recommends for an external Depth Map.
Use technical modes, but deliver the Result.
Depth, Focus, Aperture and Overlap Preview are diagnostic tools. Make the final decision in Result, in motion, under the intended color management and at delivery resolution.
Do not confuse setup speed with final quality.
Faster exists for responsive work; Enhanced/Better target different quality/performance balances. Choose final modes from a short test render because real performance depends on GPU, resolution, depth analysis and Grain processing.
Add lens character after technical reliability.
Chromatic/Spherical Aberration, Shaped Bokeh, Anamorphism and Lens Distortion can add character, but Depth Map, focus range and boundaries need to be reliable first. Otherwise styling merely hides a defect until the next frame.
08BEFORE YOU DELIVER
Validate the shot as a depth effect, not as a pretty still.
The checklist is a reminder only. A successful still cannot prove depth stability, focus animation, edge behavior or grain continuity through motion.
09QUESTIONS THAT CHANGE THE WORKFLOW
Keep depth analysis and lens styling separate.
Does AI CineFocus only work on phone footage?
No. Blackmagic uses phone footage as an obvious example of all-sharp source material, but explicitly says CineFocus can also add creative post-production control to professionally shot footage from any camera.
Is Focus Distance measured in meters to the subject?
No. The official guide states that CineFocus does not focus in physical units. It is a percentage position based on Depth Map data. The OpenFX overlay is useful for selecting a point, and the parameter can be keyframed for a focus pull.
Are Depth Quality and Bokeh Quality the same thing?
No. Depth Quality affects the internal depth-map analysis and offers Faster/Better. Bokeh Quality affects virtual defocus rendering and offers Enhanced/Better/Faster. A problem in one stage is not fixed by changing the other.
Can I use my own Depth Map?
Yes. Source supports From Alpha Input, allowing an externally prepared depth pass. Check polarity, range and routing. For an external map, the official guide recommends Separate Layers = 0.0.
Why use Remove and Replace Grain if Film Grain can be added after CineFocus?
They solve different problems. Remove and Replace Grain temporarily separates source texture so it is not blurred with the virtual defocus, then restores it. Additional Grain or a separate Film Grain effect creates new texture and can be used creatively, but does not automatically restore the source texture.
Should I always use Enhanced Bokeh Quality?
No. Enhanced makes sense when its added detail is actually visible and worth the cost. Faster/Better can be more practical during setup. Choose the final mode by comparing the real shot and a short render, not by the label alone.
Is AI CineFocus available in the free DaVinci Resolve 21?
Blackmagic lists CineFocus among the new AI tools in DaVinci Resolve Studio 21. This guide therefore treats it as a Studio feature. Check availability in your installed version and license, especially if the UI changes after updates.
Sources and the scope of this guide
Control names, mode behavior and the Studio positioning of AI CineFocus were checked against Blackmagic’s official Resolve 21 materials and the actual interface screenshots already stored in the site archive. The interactive lab is intentionally synthetic and makes no claim to Resolve’s AI model, exact parameter ranges, optical renderer or performance.
This page was updated against the current Resolve 21.1 documentation available in October 2026. UI details may change in later point releases, so the installed panel remains the final reference for exact availability and wording.