Film Stabilisation: The Complete Technical Guide
- Apr 23
- 13 min read
Updated: 3 days ago
This tutorial covers all stabilisation workflows in PFClean, from one-click automated solutions to advanced manual tracking techniques. Whether you're processing archive footage or restoring feature films, you'll find the right approach for your material. For background on the physical causes of gate weave, jitter, and instability, see our companion article here.
Quick Navigation
Introduction
High-frequency instability, jitter, weave, and shake, is one of the most visually distracting defects in film scans. Beyond viewer fatigue, it actively interferes with restoration work by making frame-to-frame analysis unreliable and manual repairs nearly impossible to track.
Stabilisation establishes a steady reference frame, allowing you to see what's actually damaged versus what's just moving around. In professional workflows, it's almost always your first corrective step.
PFClean's Workbench offers two main stabilisation tools designed for real-world archive material:
Auto Stabilize — A separate effect for quick, hands-off smoothing of high-frequency motion
Stabilize — A manual effect with three methods (Border, Area, Tracked) for greater control and geometric correction
Choose your method based on what information is available in your footage and the level of control you need.
Effect Ordering
Always stabilise first. Position your stabilisation effect at the top of the Workbench stack, before any other corrections.
Why this matters:
Manual repairs stay put. If you paint out a scratch or clone over damage after stabilisation, your fixes remain locked to the image. Stabilising afterward forces you to manually track every repair through the jitter—turning a 5-minute fix into an hour-long ordeal.
Automated tools work better. Dust, scratch, and flicker detection rely on comparing frames over time. When the image is bouncing around, legitimate detail can register as dirt, and real defects can slip through. A stable image gives these algorithms a fighting chance.
You'll see what you're actually fixing. Trying to evaluate grain, sharpness, damage or colour on a shaking image is guesswork. Lock it down first, then assess.
Note on residual motion: If stabilisation removes jitter but the image still appears to "swim" or warp, you're likely seeing film warp (buckling or dimensional instability). This requires separate geometric correction after stabilisation.
Auto Stabilize
Type: Automatic | Effect: Auto Stabilize | Complexity: Easy
When to use this
Start here for most material. Auto Stabilize analyses frame-to-frame motion across the entire image and removes high-frequency jitter without manual setup. It works best on high-volume projects where you need results quickly. Auto Stabilize subtracts camera motion from the result, focusing on high-frequency jitter. This makes it effective for both locked-off and moving shots.
Key controls
Range Slider: Sets the temporal window, how many frames before and after the current frame are used to calculate average motion. Higher values create smoother results but increase processing time.
Maximum Offset (X and Y): Prevents over-correction by capping how far any frame can be shifted. Set this just above your maximum jitter amplitude to avoid creating excessive black borders or unwanted motion.
Region of Interest (ROI panel): If only a portion of your shot is unstable you can use the ROI panel to set the in and out points of your Auto Stabilize effect.
Limitations
Auto Stabilize cannot correct rotation (gate weave) or solve for complex geometric distortion. If you see the frame twisting or warping, use the Stabilize effect instead.
Stabilize Effect Overview
The Stabilize effect offers three distinct methods, Border, Area, and Tracked, each designed for different material types and levels of control.
Correction Modes
Depending on the method you choose and the number of tracking points, you can access different correction modes:
Translation: Corrects horizontal and vertical shift
Translation + Rotation: Corrects shift and rotational instability (gate weave)
Affine: Corrects shift, rotation, scale, and skew
Perspective: Full perspective correction for complex geometric distortion
Lock Motion Setting
All three methods share the same Lock Motion setting:
Lock Motion ON: Locks the shot down completely, treating all motion as unwanted jitter.
Lock Motion OFF (default): Smooths out jitter while preserving intentional camera motion (pans, tilts, tracking moves).
Sub Pixel Motion
When enabled, Sub Pixel Motion allows the frame to be repositioned with fractional precision rather than snapping to the nearest whole pixel. This results in significantly smoother stabilization and eliminates the "chatter" often seen in high-resolution scans.
However, because this requires interpolation to recalculate the pixel grid, it can lead to a slight softening of fine details and film grain.
Enable for the steadiest possible results, especially for VFX or heavy restoration.
Disable for "purist" archival work where maintaining the original, sharp grain structure is the priority.
Stabilize - Border
Type: Semi-Automated | Method: Border | Complexity: Easy
When to use this
The Border method is the default method when you add the Stabilize effect, this is your best option when the film scan includes overscan, visible frame edges, or perforations. The frame edge should be static and unaffected by whatever is happening in the scene, making it a geometrically reliable reference for stabilisation.
Use this for:
Archival scans with full overscan
Material where the image area is heavily damaged but the edges are clean
Shots requiring rotation correction (gate weave)
How it works
A magenta border defines the tracking region around the frame perimeter. The solver locks onto the edges or perforations and calculates the necessary shift and rotation to hold them steady.
Setup:
Adjust the border width to encompass the perfs or frame line, just wide enough to capture clean contrast.
Set the Lock Motion button:
ON: Locks the shot down completely.
OFF (default): Removes camera motion from the result while stabilising the frame.
Available correction modes
The Border method can correct all modes, translation, rotation, affine, and perspective correction.
Limitations
The Border method can be thrown off when:
Motion in the scene extends close to the frame edge
Damage or debris flashes near the perforations
Sprocket holes are badly torn or inconsistent
It is a film optical which has independent edge movement
Despite these limitations, you may be surprised by what PFClean can achieve even with challenging material. Should the automated process fall short, the Area and Tracked methods provide the advanced stabilisation control needed to ensure a rock-solid result.
Stabilize - Area
Type: Manual | Method: Area | Complexity: Intermediate
When to use this
The Area method is ideal when there's no usable overscan, or when damage or in-frame motion prevents clean border tracking. Instead of relying on the frame edge, the Stabilize effect will switch to the Area method automatically when you manually draw one or more boxes over static elements within the scene, a rock wall, horizon line, building, chair, or any immovable object. This is a quick, accurate manual method that doesn't require the more hands-on complexity of individual tracker placement.
How it works
Draw rectangular areas over parts of the image that should remain stationary. PFClean automatically places multiple unsupervised trackers within each box and uses them to calculate stabilisation.
Available correction modes:
Single area: Can achieve translation, rotation, affine, and perspective correction
Multiple areas: Improve accuracy by averaging errors across tracking boxes
Error averaging: The more areas you add, the more any individual tracking errors are averaged down across the solve, presuming the area you track is a good candidate, resulting in a more stable and accurate result.
Best practices
Choose high-contrast, immovable features. Architecture, pavement, mountains, things that are physically locked in place.
Avoid moving objects. People, vehicles, foliage, reflections, specular highlights and shadows will throw off the solve.
Spread areas across the frame. For rotation and perspective correction, place boxes in different regions, not clustered in one corner.
Add more areas for challenging material. If the result isn't stable enough, adding additional tracking boxes will help average out errors and improve accuracy.
When to move to the Tracked method
The Area method works well for most shots, but when there's significant motion in the scene, even if it's just passing through your tracking boxes, the automated trackers can lose lock or become confused. Additionally, when features are only visible for part of the shot and need to be handed off between different elements, the Tracked method gives you surgical control over exactly which features are being followed and when.
Stabilize - Tracked
Type: Manual | Method: Tracked | Complexity: Advanced
When to use this
The Tracked method is for maximum control and precision. Instead of letting PFClean automatically place trackers within drawn areas (as in the Area method), you manually create and assign individual trackers from the Tracker Panel. The Stabilize effect will automatically switch to the Tracked method when you activate your trackers for the Stabilize effect in the Tracker Panel.
This is essential when:
The scene has complex motion that confuses automated tracking.
Features are only visible for part of the shot and need to be handed off between trackers.
You're working with heavily damaged material where only small clean areas are available.
You need surgical precision over exactly which elements are being tracked.
How it works
Tracker requirements for correction modes:
1 tracker: Translation (horizontal and vertical shift only)
3 trackers: Translation and rotation
4+ trackers: Affine and perspective correction
Error averaging: Just like the Area method, the more trackers you add, the more any individual tracking errors are averaged down across the solve, producing a more stable and reliable result.
Setup:
Open the Tracker Panel and create your trackers manually.
Place each tracker on a stable, high-contrast feature.
Assign the trackers to the Stabilize effect, this will happen automatically if you have the Stabilize selected.
Set Lock Motion based on your needs.
Key advantage: Non-continuous tracking
Unlike the Area method, individual trackers don't need to be active for the entire shot. You can:
Track a feature until it's occluded, then overlap with another tracker on a different feature.
Use fleeting elements, a piece of debris at the edge of frame, a brief glimpse of architecture, to stabilise specific sections.
Build a patchwork of tracking data that covers the full sequence, even when no single feature is visible throughout.
This makes the Tracked method far more flexible than the Area method for complex or damaged material. You're making the most of every small stable element in the frame, even if it's only there for a few frames.
Best practices
Track immovable objects. The same principle as the Area method, avoid anything that can move independently.
Overlap trackers generously. When one feature is about to leave the frame or become occluded, have the next tracker already established.
Use more trackers to average out errors. Additional trackers improve stability and reduce the impact of any single tracking failure.
Use F-Curves to refine. Review your tracking data and manually correct any drift or jumps.
Advanced Adjustments: F-Curves
When to use this tool
When automated or manual tracking encounters problems, extreme exposure changes, heavy damage, splice jumps or missing image data, you'll see sharp "jumps" or sliding in the stabilised result. F-Curves let you surgically correct these errors.
What F-Curves show
Stabilisation data displayed as spline graphs, plotting horizontal and vertical motion (and rotation, if applicable) over time.
How to use them
Identify errors: Look for sharp spikes, discontinuities, or obvious drift in the curve, these indicate tracking failures.
Delete bad keyframes: Remove data from frames where tracking failed. PFClean will interpolate between the remaining good data.
Smooth transitions: Manually adjust the curve to create believable motion paths where automated tracking couldn't.
Why this matters for the Tracked method
When you're overlapping multiple trackers (one feature handing off to another), F-Curves let you verify that the transition is smooth. A sudden jump in the curve means the handoff failed, either the new tracker locked onto the wrong feature, or there's a gap in coverage.
Best practice: Scrubbing through the timeline won't always reveal subtle jumps/drift that's obvious in the graph.
Combining Stabilisation Effects For A Single Clip
When to use this technique
Sometimes you need different stabilisation approaches within the same shot. A common scenario is a clip that starts locked-off, pans to a new position mid-shot, then locks off again. You want the benefits of a completely stabilised frame at the beginning and end, but smooth camera motion (without jitter) during the pan.
Where other applications take a one size fits all approach, PFClean’s Workbench workflow proves its flexibility. Because each effect in the stack processes the output of the previous effect, you can layer multiple Stabilize effects, each targeting specific frame ranges with different settings; This goes for any effects in PFClean not just the stabilisation tools.
How it works
The technique uses three separate Stabilize effects, each with a defined frame range set in the ROI panel:
Setup:
First Stabilize effect (pre-pan lock-down):
Scrub to the frame just before the pan begins.
In the ROI panel, set the effect's frame range to end on this frame.
Enable Lock Motion and apply your chosen stabilisation method (Border, Area, or Tracked).
This completely locks down the opening section of the clip.
Second Stabilize effect (post-pan lock-down):
Scrub to the frame where the pan ends.
In the ROI panel, set the effect's frame range to start on this frame.
Enable Lock Motion and apply your chosen stabilisation method.
This completely locks down the closing section of the clip.
Third Stabilize effect (smooth the pan):
In the ROI panel, set the frame range to overlap by a few frames where the pan starts and finishes.
Disable Lock Motion (turn it OFF).
Apply any stabilisation method you choose.
This removes high-frequency jitter from the camera move while preserving the pan itself.
The result
Because each effect works on the output of the previous one, you now have:
A perfectly locked-off shot at the beginning
Smooth camera motion with no high-frequency jitter during the pan
A perfectly locked-off shot at the end
Why this works
The Workbench processes effects sequentially. The first two effects stabilise the static sections completely. The third effect then processes the already-stabilised beginning and end (which pass through unchanged because they're outside its frame range) and smooths only the pan section in the middle.
Best practices
Overlap generously: When setting the frame range for the third effect, give it a few frames of overlap on either side of the pan. This ensures smooth transitions and prevents any jarring cuts between stabilised sections.
Match your methods: You can use different stabilisation methods for each effect (e.g., Border for the locked sections, Tracked for the pan), but consistency often produces cleaner results.
Troubleshooting Common Issues
Black borders after stabilisation
Cause: The image is being shifted to compensate for jitter, revealing empty space at the edges.
Solutions:
Scale the frame to remove borders using the Pan & Scan effect.
Reduce Maximum Offset (Auto Stabilize) to limit how far frames can shift.
Use the Fix Frame or the Paint tool to fill in the black area when a frame repositions significantly.
Rotation not being corrected
Cause: Insufficient trackers in the Tracked method.
Solution:
In the Area method: A single area can correct rotation, but multiple areas spread across the frame will improve accuracy.
In the Tracked method: Use at least 3 trackers to enable rotation correction mode.
Tracking fails on specific frames
Cause: Density shifts, damage, or missing image data.
Solution:
Use F-Curves to manually delete or smooth bad keyframes. The effect will interpolate across the gap.
In the Tracked method: Add additional trackers to cover the problematic section.
In the tracker panel use Blur Window, De-Flicker and update Every Frame options to help trackers hold their feature.
Border method losing lock
Cause: Motion extending to the frame edge, or damage/debris near the perforations.
Solution: Switch to the Area or Tracked method to avoid the damaged edge regions.
Area method drift in busy scenes
Cause: Movement within the tracking boxes confusing the automated trackers.
Solution: Switch to the Tracked method for manual control over exactly which features are being followed.
Unstable or jittery result even with tracking
Cause: Not enough areas or trackers to average out errors.
Solution: Add more tracking areas (Area method) or more trackers (Tracked method). Additional tracking points help average down individual errors and produce a more stable solution.
Residual "swimming" or warping after stabilisation
Cause: Film warp—physical buckling or dimensional instability in the film base that can't be corrected by rigid-body transforms.
Solution: Stabilisation (even with perspective correction mode) addresses shift, rotation, scale, and perspective, but cannot correct non-linear film warp. This requires dedicated geometric correction using the Dewarp applied as a separate effect after stabilisation.
Shot is locked down when it should show camera movement
Cause: Lock Motion is enabled.
Solution: Disable the Lock Motion radio button (default is OFF) to remove camera motion from the result while maintaining spatial relationships.
Stabilisation Quick Reference
This decision tree helps you navigate the PFClean stabilisation hierarchy to find the most efficient workflow for your specific footage.
Step 1: Identify the Primary Instability
Is it just simple gate weave or high-frequency jitter?
Use Auto Stabilize.
Do you need to remove all motion, or does the shot have rotation, scaling, or perspective shifts?
Use the Stabilize Effect and proceed to Step 2.
Step 2: Evaluate the Image Borders
Does the scan have overscan or clean, static image borders?
Use the Border Method. (Select the correction mode—Translation, Rotation, etc.—that matches the instability).
Are the edges damaged, missing overscan, or showing large jumps at the frame line?
Use the Area Method. (Track static, immovable objects within the scene).
Step 3: Assess Scene Complexity & Damage
Is there severe dirt/damage or significant in-frame motion that is confusing the automated Area boxes?
Use the Tracked Method. (Manually assign individual trackers for surgical precision).
Step 4: Final Refinement & Motion Control
Are there still minor bumps, slips, or tracking errors?
Use F-Curves to manually smooth or delete problematic keyframes.
Should the shot be completely still or preserve camera movement?
To remove all motion: Set Lock Motion to ON.
To keep pans/tilts but remove jitter: Set Lock Motion to OFF.
Does the clip contain both a static lock-off and a camera move?
Option A: Disable Lock Motion.
Option B: Combine multiple effects (using the ROI panel) for the ultimate in stability control.
See PFClean's Stabilisation in Action
Experience these workflows on your own footage. Book a live demo with our product specialists and work through your clips in real-time, or upload your material and we'll create a custom before-and-after demonstration tailored to your project.
From automated solutions to surgical precision tracking, discover how PFClean can deliver professional results on even the most challenging material.
About the Author
Adam Hawkes is a PFClean Product Specialist and restoration expert with over 20 years of hands-on experience in film and video restoration. Trained in film handling and film camera operation, Adam has contributed to more than 100 productions, including some of cinema's most celebrated titles. His expertise combines deep technical knowledge of restoration workflows with practical understanding of the physical and optical characteristics of film.
