camera
Guide layers: onion skin and reference footage in Dragonframe
Dragonframe
Definition
Dragonframe's Guide Layers panel blends the live camera feed against an already-captured frame (onion skin) or an external reference still or video, so an animator can judge a pose change or match an outside source before pressing the capture key -- distinct from traditional hand-drawn onion skinning, which stacks multiple already-drawn frames rather than comparing a live feed to a capture.
Overview
Onion skinning in traditional 2D animation stacks several already-drawn frames so the animator sees the arc of a motion at once. Dragonframe's version solves a different problem: there is no next frame yet to stack, only the puppet sitting in front of the camera right now. The software instead blends the live camera feed over a frame already captured, letting the animator see exactly how far the puppet has moved -- or needs to move -- before triggering the shutter. Live View then offers several ways to check that comparison: stepping directly from a captured frame to the live feed, an auto-toggle that switches between the two at an adjustable speed, a manual toggle key, and a punch-in magnification of the video assist image (supported on Canon, Nikon, Olympus, Fujifilm and select Sony cameras) for judging a small movement -- a finger, a mouth shape -- at native lens resolution rather than guessing from a small preview.
The same Guide Layers panel handles reference material that has nothing to do with the previous frame: an animator can load a still image or a video clip alongside the live feed, blend it in with adjustable opacity, pull a chromakey with a garbage matte to isolate just the part that matters, and pop the reference out into its own synchronized player window if it is easier to work with off to the side. That covers three different jobs with one feature -- matching a puppet's pose to an animatic or previs frame, lining a shot up against a practical plate it has to match, or working rotoscope-style over reference footage for an effects pass -- inside the same application used to shoot the animation, rather than a separate tool.
Both halves of the panel exist to replace judgment calls that used to rely on a paper exposure sheet or a director's eye on set (see exposure-sheet, on-twos): whether a move reads as smooth motion or a jump, and whether a shot lines up with what precedes and follows it, become things the software shows directly rather than things the animator has to hold in memory or check against printed notes.
The reference side is more exact than "load a clip and fade it in" suggests. Using Dragonframe 2025 lists what a reference layer can be: a movie (QuickTime, mp4, m4v, avi, wmv), a still image, an image sequence, a past take of the current scene, another Dragonframe scene, another exposure from the current take, or a secondary camera's live view. Any captured frame can also be turned into a reference layer from the timeline. The manual is blunt about what the layer is for. Reference material is never combined with the scene's output, and the chroma key is "only designed as an animation tool", because Dragonframe does not export composited sequences. The key and garbage matte exist so the animator can see through the plate to the puppet, not to make the shot.
Timing is where a reference movie can quietly mislead. A movie loaded at 100% speed plays at the speed it was shot, so a 30fps clip in a 24fps scene has to drop frames to keep that speed. ALL FRAMES shows every source frame instead, and the clip then runs at 80% speed. SPEED CHANGE and FPS are linked, a negative FPS plays the clip backwards, and HOLD FRAMES samples it on twos or threes to match the animation's own rhythm. By default the clip's first frame lines up with the scene's first frame. Dragging its thumbnails along the timeline moves that sync point, and its in and out points can be trimmed. An animator matching a plate frame for frame has to settle these settings before the first capture. Otherwise every pose is matched to the wrong instant of the plate.
The clearest documented case in this database is the "hairy leg" sequence in Kleber Mendonça Filho's *The Secret Agent* (2025), made by the Dutch studio Holy Motion (Carlos Aguilar, Cartoon Brew, 4 March 2026). Mendonça Filho shot the live-action plates first, in a park in Brazil, running the camera along 30 metres of track laid on the ground so the lens sat at the leg's height. In Arnhem, SFX propmaker Rolf te Booij built a puppet of about 30 centimetres from the life-size decomposing-leg prop used elsewhere in the film. Peter Mansfelt designed a skeleton that let the toes move separately, and he also served as director of photography. Animator Rosanne Janssens then spent three weeks animating the leg against green screen in Dragonframe, using onion skinning with the live-action plates underneath to follow the camera move. The article does not give her layer settings. In this manual's terms, though, the plate would be a reference layer blended under the live view, which is the job this panel was built for. Studio founder Mascha Halberstad called lighting the trickiest part: because the leg moves through the shot, the light has to move with it, so a matched pose is not enough on its own. The shot also needs light that matches the plate (see bluescreen-chroma-key-stop-motion).
Plate matching also depends on how fast a reference movie decodes. Dragonframe's 2027.01 development branch (build 19050, checked 2026-10-07) improves the load time of reference videos and, on macOS, adds hardware-accelerated ProRes decoding of reference layers on Apple M processors after the M1. ProRes is a common delivery format for live-action plates and edited animatics, so this bears directly on the plate-matching use above. It is development-branch software, not yet in a numbered stable release.
Workflow
- Capture a frame, then switch Live View into onion-skin mode to blend the current camera feed against that frame.
- Judge the difference -- pose, spacing, silhouette -- before moving the puppet further.
- For an external reference instead, load a still or video into the Guide Layers panel and set its opacity, or pull a chromakey and garbage matte if only part of the frame is needed.
- Pop the reference out to a separate synchronized window if it is easier to keep visible off to the side of the main capture view.
- Capture the next frame once the live feed matches the intended pose or reference position.
- For a live-action plate, load the plate as a reference layer before animating, and set its timing first: speed or FPS against the scene rate, ALL FRAMES if every plate frame must be matched, HOLD FRAMES for twos, and the timeline sync point.
- Key or garbage-matte the plate only as far as the animator needs to see the puppet against it. The final composite is made in post (see after-effects-compositing), not in Dragonframe.