Pro CamSolve puts professional camera solving inside the Unreal Editor. Footage goes in, and a tracked CineCameraActor lands in Sequencer.
Every filmmaker compositing live footage with Unreal needs a solved camera. Today that usually means leaving the engine, solving in a separate application, and carrying the result back in. Plenty of filmmakers make that work, and they find good ways to do it. I want to make it unnecessary. Camera solving should live inside the editor, next to the shot it serves, for anyone who wants it there.
A solver is the engine that recovers the camera’s path. Each one is a specialist, built for a specific kind of footage.
Follows points frame to frame, then solves with bundle adjustment. Fast and forgiving on low-texture and repetitive footage. The panel default.
COLMAP structure from motion measures the scene into a dense, metric point cloud. Slower, and the most exact. Masked movers are excluded from the solve.
A VGGT-class neural solve reads camera and depth straight from the frames, for low-texture or soft footage that feature tracking can't grip. Needs a scene without movers.
The camera turns but doesn't travel, so the solve recovers rotation only. The real focal length is the number that matters.
Confirms the shot is truly locked, then holds one camera across every frame. If it finds real motion, it points to the right solver instead.
Before anything runs, Analyze Footage samples the clip and asks two questions: how does the camera move, and is anything in frame moving on its own? It recommends the solver that fits and sets masking to match: ON for movers and OFF for clean static shots. The artist can always override it.
This is past the design phase. The full pipeline runs today: footage ingest, shot analysis that recommends a solver, lens calibration and distortion correction, SAM2 masking with prompt points, five solvers, a constraint system, point cloud review, and a one-click bake to a keyframed CineCameraActor in Sequencer.
A solved camera is the hinge between footage and everything built in the engine. These are the jobs it unlocks.
A location is shot practically and the world continues past the edge of the set. The solved camera holds the extension in place across the move, so the join stays put.
A creature, a vehicle, or a product needs to sit in footage that was shot handheld or on a dolly. The plate becomes a camera in the scene, and the element renders through it.
Scout footage becomes a working camera in minutes. A director can block a sequence against real geometry before the crew is called, and cut postvis against the plate the same day.
A phone walkthrough of a location returns a camera path and a point cloud. That cloud becomes the reference an artist builds the set against, at the right scale.
Music videos, spots, and social pieces run on schedules that do not accommodate a separate matchmove pass. Solving next to the shot keeps the work inside one session.
Students learn the concepts without a second application or a second license. The panel exposes constraints, scale, and focal length as things to reason about rather than settings to inherit.
Six principles drive the architecture. Each one removes a reason to leave the engine.
The panel, the solve, the review, and the bake live inside the Unreal Editor. A Python sidecar runs the heavy math out-of-process, so the editor never gets blocked.
All five solvers live in the same panel. Choosing one changes which controls the panel shows, and nothing else: the same footage, calibration, and bake button serve every solver.
SAM2 segmentation removes moving subjects before the solve runs, from prompt points the artist drops on the footage. Movers stop poisoning the track. Masking stays off on clean static shots, because masking a scene where nothing moves starves the solver.
A solve recovers the shape of the camera’s motion, not its size. One real measurement fixes that: pick two points in the viewer, enter the distance between them, and the whole scene snaps to metric. Fit Floor then levels the scene and sets it on the ground, so CG stands on the real floor.
Output is a keyframed CineCameraActor in Sequencer with correct scale, frame range, and focal length. Optionally, you may export an FBX through Sequencer for any other DCC.
Solvers and Unreal describe space differently, so a camera brought across can arrive mirrored or facing the wrong way. The conversion is handled in the plugin, and the baked camera matches the plate without hand correction.
These are the directions the tool grows after launch. The last two follow the research.
Exposing solve and bake parameters to Blueprints, and making solved takes available through Take Recorder, lets studios script the tool into their own workflows rather than driving it by hand.
The masking work inside Pro CamSolve is useful well beyond solving, and it already runs in 5.6. Pro Mask takes it out as its own plugin across the supported engine versions, built to work with Composure so masks generated in the editor can drive compositing directly.
A point cloud tells an artist where the world is, but nothing snaps to it. Turning the solved cloud into geometry gives set extensions a surface to sit on, shadows something to fall across, and CG a place to be positioned against rather than eyeballed. Precision Solve already produces the data this needs.
A camera solve tracks points on a world that holds still to work out how the camera moved. Object tracking flips that around. Mask everything except one rigid object, like a car, a prop, or a phone, and track points only on it. Solving those points gives the object’s motion, and pairing it with the camera solve places that motion in the scene. CG can then ride on the object: a new body on a car, a display on a screen, damage on a prop.
A sequence is rarely one shot. Queuing a folder of takes to solve unattended turns an afternoon of supervised runs into an overnight job.
Static Solve reads a shot as rigid geometry, which is why it needs a scene without movers. Feeding segmentation into the neural solve is the path to running it on shots with movers in them. This one follows the research, and the pace is set by what the open model landscape supports.
COLMAP can use CUDA for feature matching and dense reconstruction. A GPU-enabled path would cut Precision Solve runtimes on the RTX hardware most Unreal artists already own. The path is clear: a CUDA build that installs as cleanly as the CPU version does today, with licensing and redistribution sorted once for every release after it.
The first download came in December 2024. Since then the DRKR PXLS Pro Cinematography Suite has logged 5,100+ downloads across two storefronts, and Pro CamSolve is the next tool to join it. Pro CamSolve joins a product line that already has paying customers and will keep growing.
This free composition toolset introduces filmmakers to the suite with in-viewport framing guides at zero cost. Downloads here become Pro buyers.
↗ Featured on 80 Level, December 2024
This is the professional framing toolset for cinematographers working in-engine, with composition guides, custom grids, and a shared suite toolbar. It has been maintained through three engine cycles with active buyer support.
This plugin applies industry-standard filmback presets and cinema & social crops in one click. The formats are spec-accurate and mapped to the cameras working DPs shoot on, with a published reference chart.
My name is Akintayo Adewole. I trained as a mechanical engineer at Northwestern, then spent twenty years designing audio-visual systems for high-stakes rooms, including the UN's North Lawn Conference Building. I have spent fifteen years behind cameras and twenty composing and producing music.
Unreal Engine is where all of it converges. I'm currently Cinematographer / Technical Director at OVFX Animation Studios on an Unreal-built animated feature, an Unreal Authorized Instructor (Look Development), an Epic Games Animation Fellowship alum, and a Real-Time 3D mentor and course author at CG Spectrum, where I wrote the Virtual Camera Workflows specialization.
DRKR PXLS is the studio I founded in 2020, in part to build the tools I kept wishing existed on set. Pro CamSolve fills a gap I kept running into, so I went after it.
A tutorial series and documentation ship alongside the tool. Here is how I present and teach on camera today.
Matchmove without the matchmove department. Solo filmmakers and small studios get finals-capable camera solving for the price of a plugin. An entire paid external workflow stays inside Unreal.
A stronger case for Unreal as the hub. Every workflow that stays in-engine deepens Unreal's position at the center of virtual production, from previs through final pixel.
A teaching pipeline attached. As a CG Spectrum course author and mentor, I fold Pro CamSolve directly into curriculum. Learners come out the other side as working Unreal filmmakers.
A path to in-engine compositing. The SAM2 masking layer feeds holdouts and garbage mattes for compositing, with depth-based occlusion arriving in the 5.7 and 5.8 builds. The same layer seeds a future standalone matting tool in the suite.
Working prototype. All five solvers, shot analysis, lens distortion input, masking, constraints, and the Sequencer bake are operational in UE 5.6.
Footage testing + hardening. All five solvers are tested against the benchmark footage set, the lens distortion round trip is complete, and compositing against the plate lands in the panel.
Multi-version builds. The 5.6, 5.7, 5.8, and 5.9 packages go through the full QA matrix and Fab compliance review.
Fab launch. Pro CamSolve ships on Fab in Q1 2027 and joins the Pro Cinematography Suite, with documentation, a sample project, and a tutorial series live.
Pipeline integration and Pro Mask. Parameters open up to Blueprints and Take Recorder, and the masking layer ships as its own plugin.
Mesh reconstruction. The solved point cloud becomes geometry, so CG can be placed against a surface rather than a cloud.
Object tracking and batch solving. Rigid objects inside the frame get their own solve and bake path, and folders of takes can be queued to run unattended.