GLOSSARY
Rotoscoping
The frame-by-frame isolation of subjects or objects from footage — the most labor-intensive VFX task that AI has improved but not yet eliminated.
What rotoscoping actually involves
Rotoscoping (roto) is the process of manually creating mattes by tracing shapes frame-by-frame to isolate elements in footage. When a green screen wasn't used, or the key fails on complex elements like flying hair, semi-transparent fabric, or interlocking body parts, roto is the fallback. It's also used for selective color grading (isolating a face to brighten it without affecting the background), cleanup work (removing unwanted objects), and creating holdout mattes for VFX compositing. The tools have evolved significantly. Silhouette (by Boris FX) is the dedicated roto application — its per-shape animation system, layer-based workflow, and integrated paint and compositing tools make it the first choice for studios handling high-volume roto work. Mocha Pro's Roto Module uses planar tracking to automatically propagate shape keyframes, reducing manual work by 60–80% on footage with stable camera motion. DaVinci Resolve Fusion includes a solid roto-paint toolset built around its planar tracker, and After Effects has the Roto Brush (AI-assisted edge detection that works well on simple shots but falls apart on complex movement). Time estimates for professional roto are sobering. A clean human figure on a locked-off camera: 2–4 minutes per frame for a basic isolation matte. That same figure with fast movement, motion blur, and hair detail: 8–15 minutes per frame. At 24fps, a 10-second shot with complex roto needs 32–60 hours of artist time. For a walking figure with moderate motion blur across 240 frames (10 seconds at 24fps), expect 20–30 hours. This is why roto departments on feature films have 20+ artists and why AI assistance has become a competitive necessity.
Edge quality, motion blur, and common roto mistakes
The quality of roto is judged entirely by the edges. A matte that's too hard creates a cutout look — the subject looks like a sticker pasted on the background. A matte that's too soft looks blurry and imprecise. The ideal edge varies across the subject: hair needs a soft, semi-transparent edge (3–5 pixel feather), a sharp jawline needs a tight edge (0.5–1 pixel feather), and a shoulder with motion blur needs an edge that matches the blur width (which changes frame-to-frame). Motion blur is the biggest technical challenge. When an arm swings quickly, the edge isn't sharp — it's a gradient spanning 5–20 pixels of partial transparency. Your roto shape needs to follow the outer boundary of the blur, not the theoretical position of the arm, and the feather needs to match the blur width. If you trace the sharp edge and ignore the blur, you'll see a dark halo where the blur should be. Interpolation is how roto artists manage frame count. You don't roto every frame — you keyframe shapes at intervals (every 2, 4, 8, or 12 frames depending on movement speed) and let the software interpolate the in-between shapes. The faster the movement, the more keyframes you need. Fast arm movements might need keyframes every 2 frames; a slowly turning head can be keyframed every 8–12 frames. The interpolation breaks down when shapes change topology — a hand going from open to closed, hair flipping from one side to the other. These transitions need manual intervention on every frame.
Rotoscoping FAQ
Has AI replaced manual rotoscoping?
Not for professional work. AI tools (Resolve's Magic Mask, Runway ML, After Effects Roto Brush 2) handle simple, clean shots well — a person walking on a static background with no motion blur. They fail on hair detail, semi-transparent elements, fast movement with heavy blur, and occluded body parts. Professional roto still requires manual work on 40–60% of frames, but AI reduces keyframe density on the rest.
What's the right feather amount for roto edges?
It depends on the subject and the footage resolution. At 4K: skin-to-background transitions need 1–2 pixel feather; hair needs 3–6 pixels; clothing edges vary by focus — sharp focus needs 0.5–1 pixel, out-of-focus needs 4–8 pixels. Always match the feather to the actual edge softness in the plate — zoom to 200–400% and measure the gradient width at the edge.
How do I roto motion-blurred edges correctly?
Trace the outer boundary of the motion blur, not the theoretical position of the object. Set the feather to match the blur width. For a 10-pixel-wide motion blur, use a 10-pixel feather on the shape. If the blur width changes across the edge (common on rotating objects), use a variable feather or split the shape into segments with different feather values.
Need Professional Help With This?
We have a dedicated roto team that handles high-volume, high-precision isolation work for VFX studios and color grading projects — clean edges, every frame, on deadline.
Get a Free Consultation