3D animation viewer for REAPER. Load glTF and FBX animations straight onto your timeline and watch the animated character play in sync with the REAPER playhead, from any camera angle.
What's New in 0.4.0
New since 0.2.2:
Auto-Tagging (event markers): RAV finds events from the way the bones move and writes them as take or project markers. Legs: steps, lift-offs, scuffs, taps, heel taps, drags, kicks, knees, jumps and landings. Arms: grabs and releases, slaps, claps, hands on the body, punches, strikes, swings, elbows and the arms of a jump. Body: spins, jolts, sits and falls. Add any event you describe with your own rules. A doubtful event shows in amber, so you can check it by ear. See Auto-Tagging (Event Markers).
Video rendering: frame camera shots in Video view and render picture + audio with REAPER's own Render dialog. See Rendering Video.
Shortcuts window, the Model / Skeleton switch and a clip list, with the viewer options in one tabbed menu.
Installation
Requirements:
REAPER 7.x, 64-bit, Windows
A graphics card supporting OpenGL 3.3 (any GPU from the last ~10 years)
Nothing else to install. Everything the viewer needs (3D loader, UI, image decoders) is compiled inside the extension.
Already using version 0.1.x? Read Upgrading from 0.1.x first.
Install with the Demute Reaper Toolkit (easiest)
The Demute Reaper Toolkit is our tool browser for REAPER: it lists every Demute tool and installs or updates them in one click. If you already use it, skip to step 3.
Install ReaPack if you don't have it already, then restart REAPER.
Install the Toolkit (one-time setup):
Go to Extensions > ReaPack > Import repositories..., paste the following URL and click OK: https://raw.githubusercontent.com/DemuteTools/DM_ReaperToolkit/refs/heads/main/index.xml
Go to Extensions > ReaPack > Manage repositories, select Demute_Toolkit and click Browse packages.
Search for DM_ReaperToolkit, right-click it, choose Install, then click Apply.
Open the Toolkit: Actions > Show action list, search for DM_ReaperToolkit and run it. It automatically scans for available tools and updates.
Select the ReaAnimViewer card and install it.
Click Run on the card. A message confirms that ReaAnimViewer is installed.
Restart REAPER. Extensions are only loaded at startup.
Open the viewer with the action RAV: Open Viewer, or with Run on the card.
You only need step 5 once. The Toolkit downloads scripts, while ReaAnimViewer is a native extension that REAPER loads from its UserPlugins folder: the first Run copies it there. From then on, the extension updates itself.
Install with ReaPack
Install ReaPack if you don't have it already, then restart REAPER.
In REAPER, go to Extensions > ReaPack > Import repositories...
Paste the following URL and click OK: https://github.com/DemuteTools/ReaAnimViewer/raw/main/index.xml
Go to Extensions > ReaPack > Browse packages..., search for ReaAnimViewer.
Right-click ReaAnimViewer (the extension), choose Install, then click Apply. The ReaAnimViewer Launcher package is optional: it is the script used by the Demute Reaper Toolkit.
Restart REAPER. Extensions are only loaded at startup.
Check the installation: go to Actions > Show action list, search for RAV. You should see RAV: Open Viewer.
The package also installs a small RAV_Launcher script. It is used by the Toolkit; with ReaPack you don't need it (running it just opens the viewer).
Updating
With the Demute Reaper Toolkit: open the Toolkit, click Update on the ReaAnimViewer card, then restart REAPER. The extension swaps itself for the new version when REAPER closes, so the new version runs after the restart. Nothing else to run.
With ReaPack: go to Extensions > ReaPack > Synchronize packages (or wait for the automatic check), click Apply, then restart REAPER so the new version is loaded.
Manual install: download the new DLL and replace the old one, as in Manual Installation.
Upgrading from 0.1.x
Version 0.1.x came as two ReaPack packages, also when you installed it through the Toolkit. Remove the old one once:
Go to Extensions > ReaPack > Browse packages... and search for ReaAnimViewer.
Two entries are listed. Right-click the one whose category is Extensions, choose Uninstall, then click Apply. Keep the one in Scripts.
Restart REAPER, then run Extensions > ReaPack > Synchronize packages and click Apply.
Restart REAPER again.
Until you do this, ReaPack reports a conflict on reaper_animviewer.dll and ReaAnimViewer stays on 0.1.x.
Uninstalling
With ReaPack: go to Extensions > ReaPack > Browse packages..., right-click ReaAnimViewer, choose Uninstall, click Apply and restart REAPER. Don't delete the DLL by hand instead: ReaPack would still consider it installed (see Troubleshooting).
With the Demute Reaper Toolkit: uninstall the ReaAnimViewer card in the Toolkit, then close REAPER and delete reaper_animviewer.dll from the UserPlugins folder.
Manual install: close REAPER and delete reaper_animviewer.dll from the UserPlugins folder.
To find the UserPlugins folder: in REAPER, Options > Show REAPER resource path in explorer/finder (usually %APPDATA%\REAPER\UserPlugins\). Files named reaper_animviewer.dll.old or .new next to it are leftovers of an update and can be deleted too.
In REAPER, Options > Show REAPER resource path in explorer/finder shows you the right folder. Copy the DLL into its UserPlugins subfolder (usually %APPDATA%\REAPER\UserPlugins\).
Start REAPER and look for RAV: Open Viewer in the action list.
A manual install does not update itself: repeat these steps for each new version.
Troubleshooting
Getting Started
Open the viewer: Actions > Show action list, search RAV, run RAV: Open Viewer. Dock it wherever you like, REAPER remembers the position.
Drag a .glb, .gltf or .fbx file from Windows Explorer onto a track. An item is created with the exact length of the animation.
Move the playhead over the item: the character appears in the viewer, posed at that frame.
Press Play. The animation runs in sync with the transport: scrub, loop and stop as you would with a video.
Right-click and drag in the viewer to turn around the character. Start placing your sounds.
Tip: bind RAV: Open Viewer to a key or toolbar button to toggle the viewer quickly.
Supported Files
Each file should contain one skinned mesh and its animation. Animations built at runtime in the engine (state machines, blend trees, runtime IK) must be baked into the exported file. VFX, particles and engine shaders are not rendered.
Camera Controls
Model / Skeleton
The Model | Skeleton switch at the bottom left of the viewer (RAV view and Tagging view) shows the character's bones.
Skeleton fades the character and draws its bones and joints on top.
Hover a joint to see its name and the roles it plays. Click it to select it, click again to deselect. The bottom right of the view names the selected bone.
In the Tagging view, the bones the selected rule reads take the rule's colour, and + next to a condition's bone menu uses the selected bone (see Tagging your own events).
A file with no animated skeleton (a static mesh) shows No animated skeleton in this file instead.
The switch and the selection are not saved: the viewer opens in Model.
Viewer Menu
Click the menu icon in the top-left corner of the viewer to show or hide the options menu. It opens as a panel in that corner, in front of everything else in the viewer, and stays open while you tune: the render changes as you go.
On its left, a column of tabs picks which settings show:
The tab follows the view when you change view (toggle or V); otherwise the tab you picked stays. On a short viewer the settings scroll; the tabs stay in place.
Under the tabs:
Copy error log: Copy the recent load errors, the viewer version and your graphics card info to the clipboard, ready to paste into a bug report.
The installed version.
Light
Colour: Colour of the main light.
Position: Drag in the pad to move the light around the character.
Ambient: Fill light. Lower it for more contrast.
Specular: Strength of the shine on surfaces.
Relief: Strength of the normal maps (surface detail).
Ground
Enable: Show or hide the floor and grid under the character.
Grid: 1 m / 10 m: Size of one grid cell in real-world metres, whatever the file format (FBX centimetres and glTF metres are converted for you). A 100 m dragon covers 10 cells at 10 m.
Shadow
Off / Low / Mid / High: Quality of the shadow cast on the floor. Off is the lightest for your GPU.
Performance
REAPER catch-up (in the Video view tab) shows how long REAPER's Video window takes to catch up with an output size or display change while playing (see REAPER catch-up while playing).
If the viewer feels slow on a laptop, lower MSAA and Shadow first.
Shortcuts
Click the keyboard icon in the top-right corner of the viewer (left of the FPS readout) to open the Shortcuts window, a large panel over the viewer. Esc, its ✕ or a click outside it closes it. The keys work while the viewer has the focus; every other key stays REAPER's.
The window has three parts:
Sections on the left: All, then Viewer, Video, Tagging, Timelines (scroll and zoom of every RAV timeline) and Mouse. The search field at the top filters the rows by name.
The table: one line per shortcut. Action, Where it works, the Keys that run it now, what it Follows (REAPER or Own), its Source (the REAPER action it follows, or "RAV only") and the chain button.
The detail on the right, for the selected row: its keys, what it follows, its default, and anything in its way (a key another row or window takes). Its buttons: Record key…, Link… and Default.
Double-click a row to record a new key, or for a Timelines row a key or a turn of the wheel (Ctrl, Shift and Alt can be held with it; Esc cancels). Right-click a row for its menu: Record key…, Link to REAPER action…, Back to default. Mouse gestures are listed for reference (greyed, fixed).
REAPER or Own
Every row is either:
REAPER: it runs on the keys you give its action in REAPER's Actions list, read live. The viewer rows have their own RAV actions ("RAV: Viewer - …", "RAV: Video - …", "RAV: Tagging - …"); REAPER runs them too outside the viewer. The Timelines rows follow your REAPER bindings of its view actions (scroll and zoom, wheel and keys).
Own: RAV's own key (or wheel, for the Timelines rows), or a REAPER action you link the row to.
The footer says when reaper-kb.ini was last read; RAV reads it again whenever REAPER saves it.
Linking a row to any REAPER action
Click a row's chain (in Own) to link it to any action of REAPER's Main section: RAV then runs the row on that action's keys (and wheels, for a Timelines row), followed live. Find the action by:
its name (words, any order: "split edit"),
its command ID ("40757", "_SWS_ABOUT"; REAPER's Actions list: right-click > Copy selected action command ID),
a shortcut it has, typed ("Ctrl+S", "Shift+Wheel") or pressed: click the keyboard button next to the field, then press the keys or turn the wheel.
Each action in the list shows its keys. Click one to check it (its name, and the keys RAV will follow), double-click or Link to link it. Unlink (or Default) gives the row back its own key. A key another RAV row holds stays that row's.
Find by key
Find by key (top bar): click it, then press a key combination or turn the wheel with its modifiers. The window lists every action of REAPER's Main section bound to it (REAPER's defaults included) and every RAV row that runs on it and, for the plain wheel, the 3D view's fixed Zoom. Link to… on a REAPER result links a row to that action at once (the row becomes Own, linked); the footer's Undo puts the row back. Show selects a RAV row. Esc clears the result.
Default keys
A key you bind is taken from REAPER while the viewer has the focus: bind Space, for example, and Space no longer starts or stops playback from the viewer. Hold Ctrl, Shift or Alt with it to keep REAPER's own key free. The key caps and tooltips in the viewer show your keys, custom ones in the accent colour. On AZERTY and other layouts with an AltGr key, Windows treats Ctrl+Alt+key as AltGr+key, so a Ctrl+Alt binding may type a character instead.
Your keys are kept on this computer, for every project.
Working with Animation Items
Animation items behave like regular REAPER media items:
Move them to realign the animation with your sounds.
Trim the start or end to show only part of the animation.
Loop an item by extending it past its end.
Put animations on several tracks: when items overlap under the playhead, the viewer shows the one on the topmost track.
Animation items produce no audio: you can place them on the same tracks as your sounds or on a dedicated "Anim" track.
Rendering Video
ReaAnimViewer can put the animation in the videos you render from REAPER, with the session's audio, through REAPER's usual Render dialog and Region Render Matrix. Nothing is screen-recorded and RAV writes no video file: a small video FX, RAV Video FX, installed with the extension, hands REAPER the picture of its track's animation for every frame REAPER draws, in playback and in renders.
Adding the video FX
Open the viewer and click Video view at the top (or press V until it shows). The Video panel opens on the right. P, the panel button at the top right, or Video panel in the viewer menu's Video view tab show or hide it.
The panel follows the track of the animation shown in the viewer. When it says No video FX on this track, click Add video FX to track. The action RAV: Video FX - Add to selected track does the same from REAPER.
That's it. You never need the FX window: RAV drives the FX for you.
One FX shows the animation items of its own track. Where that track has no item, the FX shows nothing, so the video tracks below show through. Add the FX to every animation track you want in the video. REAPER's Video window (View > Video) shows the result too.
Video view
RAV view is your free camera. Video view is the render's camera. Switching between them never changes the free camera.
In Video view the frame shows exactly what the video will show, at the output's shape. The size and frame rate sit above it on the left, the current shot's name on the right. Nothing is drawn inside the frame.
Right-drag (orbit), middle-drag (pan), the wheel (zoom) and the navigation cube edit the shot under the playhead.
The panel's inspector edits the shot under the playhead: its name, Cut to next / Move to next, Yaw, Pitch, Distance, Target X/Y/Z, Copy RAV view (the shot takes your free camera) and Frame model (the whole model, same angle).
Track: the panel follows the track of the item shown in the viewer, or you can pin one track.
Shots
The video camera is a sequence of shots. A shot starts at a time and keeps its camera until the next shot (Cut to next), or moves smoothly to the next shot's camera (Move to next).
Shot strip (under the viewport, in Video view): the shots over the current item, and REAPER's playhead as a white line. A ruler above it shows the item's time in seconds (0:02) and, zoomed in, frames (+12f). Click or drag in it to move the playhead; double-click a shot to put the playhead on its first frame. Drag a line between two shots to retime that cut: it snaps to frames, each shot keeps at least one frame, the first shot's start stays put. Any camera move into or out of that cut speeds up or slows down to fit. Turn on Snap (next to Cut) to land drags and scrubs on the ruler's ticks; it is off by default and resets when REAPER restarts.
C (in Video view), the strip's Cut button or + Cut at playhead in the panel: a new shot starts on the frame under the playhead, with the camera shown there. Then reframe it with the mouse. A cut inside a Move to next shot makes that move end at the cut: it now eases into the camera shown at the cut instead of continuing to the next shot.
Shot list (panel): click a shot to move the playhead to it; x deletes it (the first shot stays).
Saved angles (panel): + Save keeps the camera of the shot under the playhead under a name. Click a saved angle to give it to the shot under the playhead; right-click it to delete it. Saved angles are stored in the FX, with the project.
Shots are ordinary automation: envelope points on the FX's six parameters (Yaw, Pitch, Distance, Target X/Y/Z), square points for a cut, smooth ones for a move. You can edit them in REAPER's envelope lanes, and anything that drives FX parameters (envelopes, modulation, DM-XYZ-Pad) drives the camera. Distance and target are relative to the model's size, so a shot keeps working when the animator re-exports at another scale.
Output
Size: the project's video size (File > Project settings > Video) by default, or an override for this FX: vertical, square, 4K or a custom size.
Frame rate: always the project's.
Background: the viewer's background, or Transparent (where there is no model, the picture is transparent, so the video tracks below can show through when REAPER composites the FX's transparency).
Light, floor, grid, shadow and MSAA follow the viewer's View menu (with the viewer closed, its last settings). The menu, navigation cube, FPS and shot strip never appear in the video.
Rendering
In the Video panel, click Matrix (top of the panel) to open REAPER's Region Render Matrix and tick the regions to render, or Render to open REAPER's Render to File dialog.
In the Render dialog, choose the Source (Region render matrix, Time selection, Entire project...) and a video Format (for example MP4 / H.264). Keep the video size on the project's settings, or set the size shown above the Video view frame.
Render. The picture comes from the FX at each frame's exact time, so picture and sound stay in sync and each cut lands on its frame.
RAV never changes your render settings.
REAPER catch-up while playing
While playing, REAPER prepares video frames a few seconds ahead (its read-ahead). Camera moves and playback stay in sync in REAPER's Video window. An output size or display change (light, floor, background...) waits behind that read-ahead: it can take up to that long to reach REAPER's Video window while playing. Video view in RAV is always up to date, and rendering has no offset: every frame is drawn for its own time.
While REAPER's Video window is open, the line above the Video view frame shows REAPER's current read-ahead, for example REAPER catch-up 1.5 s while playing, or REAPER catch-up: live when stopped (hover it for details). It is how long such a change can take to show there, not a measurement of a particular change. On a narrow view it shortens to catch-up 1.5 s, shortening the shot name if needed. Turn it off with REAPER catch-up in the viewer menu's Video view tab.
If the video FX shows nothing
To report a problem, click Copy error log in the viewer menu and paste it into your report.
Auto-Tagging (Event Markers)
ReaAnimViewer can find events in an animation from the way its bones move, and mark them on your timeline: footsteps, a hand hitting something, a sword swing, a body falling, a head turn, a clap... Feet, and hands that grab or let go, are found for you by Smart Detection. For anything else you can describe as "this bone goes above or below this value", you write rules (or load a preset). You check the result in the Tagging view, and commit. Each event becomes a marker named after its rule, on the item (take marker), on the timeline (project marker), or both. Then place your sounds on the markers instead of hunting for each frame by eye.
Footsteps are the ready-made starting point: under Smart Detection, tick Step and press Detect (or load the Footsteps preset, which does the same). Lift-offs, slides and pivots of the feet, and the hands' grabs, releases and pivots, work the same way. For everything else, you build your own rules, then save them as presets to reuse them.
Each animation item keeps its rules, thresholds and corrections, saved in the project. Items that play the same animation file share them, like REAPER's pooled MIDI items (see Copies of one animation share their tagging). Nothing is written to the animation file.
Quick start: footsteps
Open the viewer, then click Tagging view at the top (or press V until it shows). A strip appears under the character and a panel on the right.
Put the playhead over an animation item. In the panel's Smart Detection tab, click the Step chip to tick it, then press Detect (at the bottom of the panel).
The strip shows one row per foot (FS L, FS R) with a tick for each step found. Play or scrub to check them against the character.
Too many or too few? Drag the step's Sensitivity (higher finds more). Always a little early or late for your sounds? Drag its Offset. The markers follow as you drag. Then correct single steps if needed (see Correcting events).
As soon as you change something, RAV shows the result on the timeline as preview markers (FS L - Preview, in a darker colour). Nothing is final yet.
Select the items to tag in REAPER (the panel footer says how many), then click Commit to N items. The previews are replaced by the real markers. Cancel instead puts the items back as they were at their last Commit.
To tag several walk cycles at once, set them up one by one (steps 2 to 4), select them all and click Commit once: each item is tagged with its own rules. Copies of one cycle (the same file duplicated to build a longer walk) are set up once: they share their tagging.
Smart Detection
Smart Detection reads the body as physics: it measures the legs to know the character's size, finds where each part of the foot, and each hand, comes to rest (a floor, a stair, a ledge, a hold, a table), and places the markers from that. One setting works on every clip: there are no thresholds to tune, and no Analyse.
Every event type is a chip, grouped by the body part that makes the sound: Legs, Arms and Body (the trunk and the head). The common ones (Step, Lift-off, Jump, Land, Grab, Release) come first, in bold. Click a chip's name to tick or untick the type; click its ⚙ to see its settings in the strip under the list (an unticked type is ticked at the same time; its ⚙ again hides them). A category's caption shows how many of its types are ticked (2 / 11); its × unticks them all. The chip list takes the tab's free height and is the only part that scrolls. The settings strip keeps one compact height, so nothing moves while you tick, search or switch types: its header (the type's tick, its category, (i) with its whole description, × to hide it), then the type's description in the lines its settings leave free (an amber note, such as "Raised by Detect on FS R", comes first), Sensitivity, Offset and Per side, and the type's own options. A type that needs more room (Per side open on a type with many rows) scrolls inside the strip. An amber dot on a ⚙: Detect raised that type's sensitivity, or found nothing up to 100 %; the tab shows an amber dot too while the Rules tab is open.
Is looping (beside the search, saved on the item): the animation loops, its end continues into its start, so a contact, step, grab or release that crosses the loop point is found whole.
Search fades the chips that do not match (by name, marker, category or description) without moving them.
Presets are your own sets of ticks, kept on your PC for every project. + Preset saves the types ticked now under a name; clicking a preset ticks exactly its types and unticks the others (press Detect to apply them); a preset shows a tick when the ticks are exactly its own. Its ⚙: Rename..., Store current ticks (it takes the ticks of now), Delete....
Tick the event types you want, then press Detect (one button for every category):
Legs
Step: where the foot lands. Combined gives one marker per landing (FS L, FS R), at the first part of the foot to touch. Separate gives the heel and the toe their own markers (Heel L, Toe L, Heel R, Toe R). The toe is the ball of the foot (the toe bone), or the toe tip on a rig without a toe bone. A foot that glides into place (high heels on a catwalk) gets its step where it stops.
Lift-off: where the foot leaves the ground (Lift L, Lift R). Combined gives one marker per lift-off, where the heel and the ball have both left the ground (the toe tip, which rolls off a little later, counts only on a rig without a ball bone). Separate puts each one on the block of that part: Lift Heel L for the heel, Lift Toe L for the ball or the toe tip.
Slide scuff: where a planted foot slides on the ground (Slide L, Slide R).
Pivot scuff: where a planted foot turns on its ball or its heel (Pivot L, Pivot R): flat, on tiptoe with the heel up, or on its heel with the toes up. A foot that turns as it rolls onto its toes and stays there (a swivel) pivots; so does a wide turn that ends as the foot leaves the ground (a spin turn, of 30° or more by default; a higher pivot sensitivity lowers that angle); a small twist as the foot leaves (a heel whip at toe-off) does not.
Start / Peak on both scuffs (and the hands' Pivot scuff): the marker where the scuff starts (Start, the default) or where the part on the floor rubs fastest (Peak: where a dance's swivel squeaks).
Jump: where a jump takes off, on the last foot to leave the ground (Jump L, Jump R: one marker per jump; when both feet leave together, within 10 ms, it is Jump L). A jump is found in three ways: by its flight, both feet off the ground long enough (300 ms by default) while the body falls freely; by a hard push off the ground, the hips accelerating up well above body weight while a foot is down, with both feet leaving right after (even briefly); or by a free fall, both feet off as long as a flight and the body falling freely over its first 300 ms, however it ends (a jump that ends on the back or the side gets its Jump, and no Land). Feet off the ground while the hands hold the body up (hanging, climbing) give none. A run's strides push as hard as a jump, so a push counts only while the body is not running (about 2 m/s); a jump taken while running is found by its flight. A run's own flights (both feet off at every stride, longer as it speeds up into a sprint) are never jumps, whatever the sensitivity: a flight counts as a jump only when it stands out from the flights just before and after it (2.5 times longer), so a jump taken mid-run is still found. A leap from one foot onto the other, like the Jumping sample onto a ledge, gives its Jump by its push. A clip that starts during the push (the take-off foot already on the ground at the first frame) gets its Jump too, unless Is looping is on. A higher sensitivity counts shorter flights and softer pushes. Separate: on the part that leaves last, of the heel and the ball as for Lift-off (Jump Heel L, or Jump Toe L for the ball, or the toe tip on a rig without a ball bone). With Lift-off ticked too, Jump replaces the lift-offs within 30 ms of it (a Jump Heel / Jump Toe only the lift-offs of its own part, when lift-offs are Separate too).
Land: where the jump lands, on the first foot to touch, at that foot's step (Land L, Land R; both feet together, within 40 ms: Land L). When the second foot touches 40 to 90 ms after the first, the two impacts are heard apart: each foot gets its own Land (Land L, then Land R at its own step); later than that, the second foot's touch is a step (a running stride never comes that fast). A take-off keeps one Jump: the foot that leaves first is a lift-off. Separate: on the part that touches first (Land Heel L, Land Toe L). A foot that lands sliding fast along the floor holds no contact, so it gives no Land. With Step ticked too, Land replaces the steps (FS, Heel, Toe, either foot) within 30 ms of it, and the second foot's step of a landing on both feet together (so a landing is one Land, or two, never a Land and a step side by side); a Land Heel / Land Toe replaces only the Heel / Toe steps (or FS), so the heel slap that follows a toe landing keeps its own marker.
Tap: where the toes touch the ground lightly while the other foot carries the body (Tap L, Tap R): a touch of the ball or the toe tip under 150 ms, the heel up, the other foot on the same floor. Steps of a walk or a run are never taps. A higher sensitivity counts longer touches (up to 300 ms). With Step or Lift-off ticked too, Tap replaces that foot's step at the touch (FS or Toe) and its lift at the end of the touch: a tap is one sound. Detect never raises Tap or Drag on a clip that has none.
Drag: where the toes slide along the ground while the foot moves, the heel up (Drag L, Drag R): a foot dragged behind, the feet of a knee crawl, the back foot of a kneel. The marker sits where the slide starts. A walk's or a run's toes skimming the floor as they leave or land are no drag; a foot moving fast (about 2 m/s) is swinging, not dragging. A higher sensitivity counts shorter drags (100 ms by default).
Heel Tap: where the heel drops back onto the floor while the ball of the foot stays down (Heel Tap L, Heel Tap R): the foot rolls onto its ball, the heel rising a centimetre or more, then comes down, as in a dance. The marker sits where the heel reaches the floor. A foot lifted flat, a foot on a climbing hold or a crawl's toes give none; a heel kicked up and slammed down fast enough to leave the ground is a Step (Heel), not a heel tap. A higher sensitivity counts smaller heel rises. With Step ticked too, Heel Tap replaces that foot's step at the same moment. Detect never raises it on a clip that has none (a weight shift lifts the heel a little).
Knee: where a knee touches the ground, kneeling or crawling (Knee L, Knee R): the knee comes down and rests close to the floor, or comes down onto the floor and slides on (a low crawl dragging its knees). A knee that stops a little above the floor (a deep crouch), or legs that stop sliding along the floor, give none.
Kick: where a foot whips through the air on a raised, stretched leg (Kick L, Kick R): its speed against the body's, at its fastest (Peak) or where the motion starts (Start). A running stride's foot moves as fast, but it is folded under the hips when it is up and low when the leg is stretched, so runs give no kick. Separate: on the part fastest at the peak (Kick Heel L, Kick Toe L).
Arms
Grab: where the hand lands on something and stays: a hold, a ledge, a table, the floor (Grab L, Grab R). A hand that only slows down, or stops for an instant between two moves, is no grab.
Release: where the hand leaves what it held (Release L, Release R).
Pivot scuff: where a planted hand turns on what it rests on, a hand on the floor or a ledge while the body swings round it (Hand Pivot L, Hand Pivot R). The turn is the hand's own twist: a vault where the palm stays put gives no marker. A hand that lands turning, or leaves at once, does not pivot. Only after a Grab (on by default): only a hand that came down with a grab (or rests from the start of the clip) can pivot. Off: any resting hand can, a hand placed gently too, but a hand that twists while held still in the air can then give an extra marker.
Elbow: where an elbow touches the ground, lying or crawling on the elbows (Elbow L, Elbow R): the elbow comes down and rests on the floor.
Clap: where the two hands meet at speed and stop against each other (Clap L, Clap R: one marker per clap, on the hand that moves faster). Hands that pass each other, or come together slowly, give none. A higher sensitivity counts slower claps, and hands meeting a little further apart; Detect never raises it by itself, so a clip without a clap stays without one. Two hands joining on a two-handed grip (a pistol, a rifle) meet as a clap does and give one.
Hand on body: where a hand lands at speed on the body, the trunk, the head or a thigh, and stops there (Hand on body L, Hand on body R). A hand that swings past the body, or comes to rest slowly, gives none, nor does the body coming onto a hand held still (a climb's last pull); a hand grabbing gear worn on the body (a holster) is no touch. A higher sensitivity counts slower touches, a little further from the bones; Detect never raises it by itself, so a clip without a touch stays without one.
Slap: where a flat hand hits something, the floor, a wall or the body, and the fingers stay straight (Slap L, Slap R): a hand crawling on the floor, landing on its palms, slapping a thigh. A hand whose fingers close after the contact (a hold, a rung, a pistol grip) is a Grab, and a fist is never a slap. With Grab or Hand on body ticked, a slap replaces that hand's grab or hand on body: one marker per contact. It reads the fingers, so it needs finger bones that move in the animation: on a rig without fingers, or whose fingers do not move around the contact (within half a second of it), it finds nothing and the contacts stay grabs. A contact in the last 0.15 s of an animation that does not loop stays a grab (the fingers can't be seen after it). Detect never raises it by itself.
Swing: where a hand whips through the air (Swing L, Swing R): its speed against the body's, at its fastest (Peak) or where the motion starts (Start). A hand resting on something gives none. The default catches strikes and big throws of the arm; the arms of a walk or a run stay below it, but swinging the arms into a jump can reach it. A swing, kick or punch already under way on the first frame (begun before the animation) gives none, unless Is looping is on. A straight thrust away from the shoulder is a Punch, not a swing; a fast arc blow that stops sharply is a Strike.
Punch: where a hand thrusts straight out from the shoulder (Punch L, Punch R), at its fastest or at its start, as Swing: the arm opens from bent to straight, fast, at about shoulder height or a little lower. A hook or an uppercut (an arc, the arm already open) is a Strike; an arm thrown straight down (to the floor, for balance) or reaching up (a climb's reach for the next hold) is no punch. A higher sensitivity counts slower punches.
Jump and Land (Arms): the arms of a jump. Jump: where an arm is thrown up into the take-off (Arms Jump L, Arms Jump R); Land: where an arm comes down at the landing, reaching for the floor or the knees (Arms Land L, Arms Land R); both at the hand's fastest. They follow the take-offs and landings that Legs > Jump / Land find at their default settings (ticked or not; changing their sensitivity does not move the arms), so an animation with no jump has none; a landing on both feet apart (two Lands) still gives one Arms Land per arm. With Swing ticked, they take that arm movement from it: the swing gives way, one motion, one sound, and Swing keeps the swings that have nothing to do with a jump. A hand that hits the floor is still a Grab. A higher sensitivity counts slower arms.
Strike: where a fast arc blow of the hand stops sharply (Strike L, Strike R), at the hand's fastest: a hammer fist, a hook, an uppercut, a sword or axe blow. The blow either hits (the hand stops dead and stays stopped: a blade into the floor; a hand that comes straight back is no hit) or travels an arc in front of the body, the arm raised and the hand moving forward, and stops within about 0.2 s. A straight thrust is a Punch (Punch wins on the same blow); an arm that swings on after its peak (a follow-through, a dance move), an arm pulled back (the other arm of a blow), a hand that snaps back the way it came (a feint, a dance move), a running arm and an arm thrown up into a jump (Arms > Jump) are no strike. With Swing ticked, a strike takes that hand's swing (one motion, one sound); with Arms > Land ticked, that hand's landing arm at the same moment (a blow landing with a jump). Whether the blow is a fist or a weapon cannot be read from the bones. A higher sensitivity counts slower blows; Detect never raises it by itself.
Body
Jolt: where an impact jolts the whole body, for gear and cloth layers (one Jolt marker, no side): a step, a jump's landing, a knee coming down, the body sitting down or a fall landing on the ground (as Knee, Sit and Fall find them, ticked or not), where the body (the hips) is thrown up hard, 1 g or more by default, which means the ground pushes with about twice the body's weight. Runs, sprints and landings give one; ordinary walking steps do not. Raise the sensitivity to catch softer steps (0.5 g at 100 %), lower it to keep only the hardest landings (2 g at 0 %). A landing and its steps make one jolt. A clip with no jolt (a walk, a climb) stays without one: Detect does not raise this sensitivity. The inspector's Strength is the peak, in g.
Spin: where the whole body turns round, a turn on the spot, a dance spin, a turn while walking (Spin where the turn starts, Spin End where it stops; no side). It is read from the hips: a turn starts when they turn faster than 15° per second, ends when they slow under 5° per second or turn back, and counts when it turns 45° or more, so a walk's hip sway gives none. A turn already under way when the clip starts, or still going when it ends, gets its marker at the clip's edge. A higher sensitivity counts smaller turns (22° at 100 %); Detect never raises it by itself, so a clip without a turn stays without a Spin. The pelvis twisting while the hands hold on (climbing) can give a Spin of about 50°. The hips turning out and straight back (a punch's twist) give none: two turns back to back, opposite ways and of similar size, are a twist.
Sit: where the body sits down, on a seat or on the floor (Sit, one marker for the body): the hips come to rest low, the back upright and the hips away from the feet. A squat (the hips over the feet), a kneel (a knee down) or lying down gives none.
Fall: where the body lands on the ground, on the hips or the back, after a fall (Fall): the hips or the back come close to the floor after falling fast, such as a fall on the back after a free fall. Lying down slowly gives none, nor does a squat, however deep and fast (the back upright over the feet); a higher sensitivity counts slower falls.
Detect lights up while a change waits for it: a type ticked or unticked, a type switched between Combined and Separate, a whoosh (Kick, Swing, Punch) or a scuff switched between Peak and Start, or the hands' Only after a Grab switched. Your corrections follow a switch: from Combined they go to the heel's row, from Separate back to the combined row of that foot. Each ticked type adds one row per foot (or per hand; one row for Jolt, two for Spin) to the strip, next to the rules' rows.
Sensitivity (%, one per type): higher finds more events, lower fewer; 50 % is the default. Offset (ms, one per type) moves every marker of that type, for example +20 ms if your sound should land after the contact. On a type already detected, both apply at once: the markers follow while you drag.
Click an event of a Smart Detection type on the strip: the Smart Detection tab opens and its settings strip shows the event instead (Event mode, same place): < Jump > Event Jump L, Go to event, Suppress (or Delete / Restore), its State, Time, Strength (the foot's or hand's speed, the turn's angle, the jolt's g), Speed, Found by and Confidence; < Jump goes back to the type's settings, × closes it. An event of a rule opens the Rules tab instead.
Marker names stay plain (FS L): the strength of an event shows in its Event view only. Choosing a sound by strength will come with a later version.
Doubtful events (amber): each event Smart Detection finds gets a confidence: how much stricter its thresholds can get before it disappears (100 %: still found with thresholds twice as strict; 0 %: only just over one; for Punch, 1.4 times as strict, a jab's speed sitting close over its threshold). Below the limit set in the viewer menu, Tagging > Doubtful events > Amber below (40 % by default; 0 % turns it off), the event shows in amber with a ? in the strip, its inspector says why, and its marker is written in amber (its name unchanged): check those by ear first. The limit is one setting for every type, kept across sessions; changing it shows previews until the next Commit. The confidence only knows the thresholds the sensitivity moves: an event that clearly passes them but is not the motion the clip means (a deep crouch whose knee stays just off the floor, two hands joining on a pistol grip read as a clap) is not amber.
Kick, Swing, Punch and the arms' Jump and Land are never raised by Detect when an item has none: most animations have no whoosh, and raising them would read a run's arms as swings. Raise their sensitivity yourself to catch slower motions.
Correct single events as on any row (see Correcting events): your corrections survive a new Detect and a sensitivity change. After an offset change, check them: a suppressed event stays masked only while the detection is within 30 ms of where it was, and your own events stay where you put them. Unticking a type removes its rows and the corrections on them.
Click a row's name in the strip to select it: the Smart Detection tab shows its type's settings. Its marker name and colour are in the Rules tab's inspector (switch to that tab after clicking the row), where you can change them.
Roles: Smart Detection needs, for at least one leg, the heel, the knee and the hip (up leg; the hips will do), or else the knee and the hip (up leg) alone; per foot the heel and the toe (or the toe end); per hand its hand bone; for Jump, Land and Jolt, the hips bone plus a heel, toe or toe end on both feet; for Spin, both hips (up legs); for Tap, Drag and Heel Tap, the heel and the toe (or the toe end) of the foot; for Knee and Elbow, that knee or elbow; for Sit, the hips and the chest (or spine); for Fall, the hips or the chest (or spine); for Clap, both hands; for Hand on body, its hand and two body bones in a row (the hips and the spine, or the hips and an up leg...); for Slap, its hand, the knuckle (the base of a finger: the middle one, else the index, else the ring finger) and the fingertip of that finger. Sit and Fall also need a foot bone, to find the floor. Kick, Swing, Punch and Strike also need the chest (or the spine, or the hips); the arms' Jump and Land need that and what Jump and Land need. When a ticked type misses one, the panel names it in red (at the top of the chip list) and that foot or hand gets no events of that type: map the bone in Roles. A rig without foot bones still gets its hand events, as long as the knee and the hip (up leg) of one leg are mapped.
Smart Detection and rules work together on one item: each has its tab (the Rules tab lists only the rules), and both make markers. A preset saves both.
Tagging your own events
In the Tagging view, put the playhead over the item and click + Rule. Give the rule the name you want on the markers (for example Whoosh R) and a colour.
Set its condition: pick the bone to watch, what to measure (position, speed, acceleration, stillness, relative drop), in which direction (vertical, horizontal, total...), and from what (the floor or another bone). The bone menu lists the roles first, then every bone of the skeleton, including a weapon or prop bone if it is part of the rig. Not sure which bone is which? Switch the 3D view to Skeleton (bottom left), click the bone on the character, then click + next to the condition's bone menu: it uses that bone, or its role when the bone plays one. When it plays several, it uses the role the bone is named for (ball_l is the toe, not the toe end), else the bone.
Watch the signal in the strip while you scrub, and drag the threshold line to where the event happens. Add more conditions with + AND condition if one is not enough.
Choose where the marker lands: at the start (the moment the threshold is crossed) or at the highest / lowest point of a signal (for example the fastest moment of a swing).
Happy with it? + Save as... in the preset menu, and load it on your other clips.
A few ideas to start from (thresholds depend on your character and animation, use Analyse or drag the lines):
Rules
A rule produces one kind of marker. It says when an event happens (its conditions), where the marker lands, and how often it may fire. Click a rule in the panel's Rules list to see it in the inspector below.
Rules list: the switch turns a rule on or off, the number is how many markers it makes on this item, and the two icons Duplicate and Delete it. + Rule adds an empty one.
Name and colour: the name is the marker's name on the timeline; the colour dot sets the marker's colour (project markers) and the rule's colour in the strip.
When all of these hold: the rule's conditions. They all have to be true at the same time. Each condition reads like a sentence of small menus: IF [Bone / Joint angle / Rotation] [which bone] [Position / Speed / Acceleration / Stillness / Relative drop] [vertical, horizontal, total, X, Y, Z] from [the floor, or another bone] goes [below / above] a threshold, with a margin. + AND condition adds one; the cross removes one.
Stillness is how far the bone moves over the next 150 ms, in cm (in ° on an angle). "Stillness below 1 cm" finds the moment a hand or foot settles, even if the motion capture jitters. It comes true slightly before the contact; if it is always early by the same amount, an offset fixes that. On an animation exported in place (the character does not move forward), a planted foot slides back: use vertical.
Relative drop is the bone's speed as a share of its fastest speed over the 300 ms before, in %. "Relative drop below 20 %" finds a stop whether the approach was fast or slow. On one direction only (vertical, X...), the speed also drops where the motion turns back, at the top of a lift: use total, or add a height condition.
After you change a condition's measure, click Analyse: the threshold keeps its old number, which means something else for the new measure.
Bone: a point on the character, for example the left heel's height above the floor.
Joint angle: the angle at a joint, such as the knee (180 = straight leg).
Rotation: how much a bone turns, relative to its parent or to the world.
Threshold and margin: the event fires when the signal crosses the threshold. It can only fire again after the signal has come back past the threshold by more than the margin (like a gate's hysteresis), so a shaky foot does not give a burst of steps.
The lock icon: a locked (fixed) threshold is never changed by Analyse.
Place the marker:at the start of the match, or at the highest point / at the lowest point of one of the rule's signals. The match is the stretch where all the conditions hold. A contact is best marked at the start (the moment a hand reaches the table); a swing is best marked at the highest point of its speed.
Offset, Min length, Cooldown (in ms):
Offset moves every marker of the rule, for example +30 ms if your sound should land after the contact.
Min length: the match must last at least this long to count.
Cooldown: no new marker from this rule until this time has passed.
Item options (the gear icon next to Roles):
Sensitivity drops events weaker than this share of the clip's strongest (0 = off).
Edge margin ignores events this close to the clip's start and end (handy for T-pose frames).
Number fields have no sliders: drag sideways to change them (hold Shift for fine steps), or click and type a value.
Presets
A preset is a set of rules you can load on any item. The Preset: field at the top of the panel shows the item's preset; click it to open the preset menu.
Factory presets come with ReaAnimViewer and update with it. They are read-only. One ships today:
Footsteps: Smart Detection of steps, combined: an FS L / FS R marker each time a foot lands.
Items set up with an earlier Footsteps Heel show Legacy · v3 is installed: Update switches them to Smart Detection (their corrections stay; those made on the old heel rules may show as orphan suppressions or your own events to delete), Keep v2 keeps their rules. Footsteps Toe is gone: items set up with it keep their rules, and the preset field says the preset is gone.
User presets are yours. They live in your own folder, which updates never touch, so you keep them when ReaAnimViewer updates. The bottom of the menu shows that folder, with a button to copy its path.
Using the menu: type in the search field, or hover Factory or User to see the list. Double-click a preset (or select it and press Enter) to load it on the item under the playhead. Right-click a preset for Load, Rename... (F2), Delete (Del) and Export....
Save (it shows the preset's name) writes the item's current rules over its user preset. + Save as... writes them to a new user preset: type a name and press Enter. A factory preset can't be overwritten: use Save as. Overwriting or deleting a preset asks first, inside the menu (in amber): click Overwrite or Delete (or press Enter) to confirm, Cancel (or Esc) to keep it. A deleted or overwritten preset can't be brought back with Ctrl+Z.
Import... copies a .ravpreset file into your User presets. Export... writes the selected (or loaded) preset to a file you can send to a colleague.
What an item keeps. An item keeps its own copy of the preset it was set up with. Changing a preset never changes your items behind your back:
edited next to the preset name means you changed this item's rules or thresholds since loading the preset. That is normal: your thresholds belong to the item. Save or + Save as... to turn your changes into a preset.
Legacy · v3 is installed (an amber band) means the preset was updated since this item copied it, for example by a ReaAnimViewer update. Click Update to take the new version (it replaces the item's rules), or Keep v2 to stay on the version you have. After Keep, the preset name shows kept v2 and the band does not come back for that version.
Loading another preset on an item replaces its rules and starts its event list over (your corrections on that item are dropped).
Roles
Presets don't name bones: they name roles. RAV's built-in roles come in three groups:
Legs: left and right heel, toe, toe end, knee and hip (up leg).
Arms: left and right hand, elbow, shoulder (upper arm) and clavicle.
Body: hips, spine, chest, neck and head.
A role is a point: the joint where its bone starts, as Skeleton mode draws it. So the knee is the shin bone (LeftLeg, calf_l), the elbow is the forearm (LeftForeArm, lowerarm_l), the shoulder is the upper arm (LeftArm, upperarm_l), and Mixamo's LeftShoulder (clavicle_l) is the clavicle. When you load a preset, RAV finds the bone for each role from the bone names. It knows Mixamo names (mixamorig:LeftFoot, LeftToeBase, LeftHand...) and the names of Unreal's mannequins (foot_l, ball_l, hand_l, spine_05...). When a rig lacks a role's own bone, RAV takes the nearest bone of the same chain: the mannequins have no toe end bone, so their toe (ball_l) plays the toe end too, and a rig without Spine2 (or spine_05, spine_03) gets its highest spine bone as the chest. A rig without toe ends gets its toe base the same way. The mapping is per skeleton: every animation of the same rig gets the same bones.
Roles ✓ (top of the panel) means every role the rules need was found on this skeleton.
Roles · 2 missing (in red) means some roles have no bone on this skeleton. Hover it to see which ones. Only the rules that read them are skipped: each one's row in the strip says No bone for: ... in red, its events can't be edited, and its markers stay as the last Commit wrote them (a rule never committed has none). The other rules and Smart Detection still run, show their previews and are committed as usual. The same goes for an Smart Detection type whose own part has no bone (for example Grab with no left hand bone): its earlier markers are kept, never removed. A missing bone never deletes markers already placed. When nothing else can run (every rule that is on misses a role and no Smart Detection type is ticked), the strip shows a message instead of the signals and Commit skips the item ("skipped: roles"). Mapping the role brings its rule back at once. Analyse still needs every role.
Click Roles to open the Skeleton & roles window. It lists every role (under Legs, Arms, Body, then Your roles), the bone that plays each one, and where that bone comes from: auto (RAV's guess from the names), you (your choice) or none. The title says "all found" or how many roles have no bone. The ? next to the undo arrow sums up the window; Esc or the cross closes it.
To fix a role, open its menu and pick a bone (type in the search field to find it), — none — for no bone, or Auto (...) to go back to RAV's guess. The strip, Roles and Commit follow at once.
Your choice is remembered for every item with the same skeleton, in every project (it is saved in REAPER's resource folder, ReaAnimViewer/roles.txt, not in the project). Role changes are saved at once and are not REAPER undo points: while the window is open, Ctrl+Z (or the undo arrow next to the close button) undoes them one by one. After you close it, pick Auto (...) to go back to the guess.
You can also pick a raw bone in a condition's bone menu: the Bones list under the roles has every bone of the skeleton.
Your own roles. For anything the built-in roles miss (a weapon tip, a tail, a finger), type a name in the Skeleton & roles window and click + Role (or press Enter): "Sword Tip" becomes the role sword tip, listed for every project. Map it per skeleton like the others, then pick it in a condition's bone menu (it is listed under Your roles, after the built-in ones). A preset that uses it works on any rig where the role is mapped: on a new rig, RAV guesses its bone from the bones you picked for it on other rigs (same name, any namespace), shown as auto.
Right-click a role of yours (or hover it and press F2) to rename it: its rules, presets and bones keep working. Its ✕ deletes it (confirmed in place): its bones stay remembered, so creating it again brings them back, and rules that use it keep it. A role read by a rule but not in your list (from a colleague's preset) shows as "not in your roles" and can still be mapped. Creating, renaming and deleting are undone with Ctrl+Z like a bone change while the window is open.
A role of yours made before its name became a built-in role (for example left hand) becomes that built-in role: rules and presets keep working, and so do the bones you picked for it. It leaves your list and takes the built-in name.
Share your roles.Export... in the Skeleton & roles window writes your roles and this skeleton's bones to a .csv file that opens in a spreadsheet or a text editor. Pick comma or semicolon separated in the save dialog (semicolon suits Excel in French and other regions). The proposed name is RAV_Roles_: add the rig's name after it, since roles belong to a rig, not to one animation. Import... reads such a file, yours or one edited by hand, into your roles for the skeleton of the item it is opened on: missing roles are added, the file's bones replace yours, and rows whose bone this skeleton lacks are skipped and counted. Ctrl+Z undoes an import. Role names cannot contain , or ;.
The Tagging view
V cycles through RAV view, Tagging view and Video view, or click one at the top of the viewer. The Tagging view uses the same camera as RAV view.
The strip (under the character) shows the item under the playhead:
One row per rule, in its colour, with a tick for each event. Your own events are labelled you; suppressed ones are faded.
Below the rows, the selected rule's signals, one lane per condition, with the threshold as a line and the margin's re-arm level as a dashed line. Drag either line to tune it: the markers follow live.
Click or drag to move the playhead. Alt+wheel zooms, Shift+wheel scrolls. Drag the strip's top edge to make it taller (more room for the signal lanes).
Analyse proposes this item's thresholds from the clip. It only runs when you click it, and it leaves locked conditions alone.
+ Event (key E) adds your own event at the playhead on the selected rule.
The 3D view has the Model | Skeleton switch at the bottom left (see Model / Skeleton). In Skeleton, the bones the selected rule reads are drawn in the rule's colour, and + next to a condition's bone menu puts the selected bone in it (with no bone selected, + switches the view to Skeleton).
The panel (on the right, drag its edge to resize it): the item's name, Roles, the item options, the Preset: field, then two tabs of fixed width, each taking the whole height down to the footer:
Smart Detection · N on: the presets, the search and Is looping, the type chips (the part that scrolls), then the settings strip (see Smart Detection).
Rules · N: + Rule, the rules list (switch, colour, events, name, Duplicate, Delete; it scrolls past 6 rules), then the inspector for the selected rule or event of a rule (it scrolls). Missing roles are said at its top.
An amber dot on a tab that is not open: it has something to tell you (a raised sensitivity, a missing role).
The footer stays at the bottom, two lines that never move:
Detect · Commit to N items · Cancel.
The status: the selection, preview not committed (or markers up to date) and the last Commit; what does not fit drops off the line and stays in its tooltip. Then ⋯: Commit writesTake markers, Project markers or Take + project markers (for every project), and Clear item data... (it asks first, in a bubble).
Correcting events
Detection is never perfect. Fix single events directly in the strip; the rules stay as they are.
Click an event to select it. The inspector shows its time, strength and speed.
Drag an event to a new time: it becomes yours (your own event, at a fixed time). Detection never moves your events, even when you change the thresholds later.
Double-click a rule's row to add your own event there, or press E to add one at the playhead.
Right-click an event, or select it and press Delete:
a detected event is suppressed: it is masked and gets no marker (any detection of that rule within 30 ms of that time stays masked);
a suppressed event is restored;
your own event is deleted.
The inspector's button does the same (Suppress, Restore or Delete). Go to event moves the playhead to it.
When the animator delivers a new animation on the same item, RAV detects it again from scratch and your corrections on that item are dropped.
Markers
Which markers: in the viewer menu's Tagging view tab, Markers written: Take (markers on the item), Project (markers on the timeline, in the rule's colour) or Both (the default). The panel footer's ⋯ sets it too (Commit writes). It applies to every project.
Preview, Commit and Cancel:
Every change you make (a threshold, a rule, a preset, a correction) is shown at once as preview markers, named <marker> - Preview, in a darker colour, next to the markers you committed. They disappear when the result is the same as the committed one.
Commit to N items writes the real markers of every selected item that has rules, each with its own rules, and removes their previews. Items linked to a selected one (see Copies of one animation share their tagging) are committed too: the footer counts them as "linked". It replaces only the markers RAV wrote: your own markers are never touched. If one of your markers already has the same name at the same time, RAV leaves it and writes no duplicate. Items without rules are skipped and the footer says so.
Cancel removes the previews of the selected items (and of the items linked to them) and puts their rules, thresholds and events back as they were at their last Commit.
The button reads Nothing to commit when the selected items' markers are up to date. The footer line says markers up to date, preview not committed or markers not written yet.
Every change, Commit and Cancel included, is saved in the project at once and is one step in REAPER's undo history. There is no separate "save" step.
Markers follow their item
RAV's project markers stay attached to their item, even with the RAV window closed:
Move or trim the item: its markers move with it. A marker whose event falls outside the trimmed item disappears, and comes back when you extend the item again.
Split an item: each part gets the markers of its own part.
Duplicate or copy-paste an item: the copy gets its own markers (and shares the original's tagging, see below).
Delete the item: its markers are deleted too.
Take markers always follow their item, as in any REAPER item.
Editing the markers in REAPER
You can also correct events right on REAPER's timeline, with the RAV window open or closed. The item's event list follows your edit:
Drag a RAV marker (take or project): it becomes your own event at the new time, as if you had dragged it in the strip.
Delete a RAV marker: its event is suppressed (or removed, if it was your own event).
Rename a RAV marker, or drag it outside its item: the marker becomes yours. RAV leaves it where it is, never moves or deletes it, and suppresses its event.
With Both, the other marker of the same event follows (drag one, the other moves; delete one, the other goes).
After editing a committed marker, the item still reads markers up to date: no preview appears. Ctrl+Z undoes the marker edit and the event change together.
The items linked to it follow too (see the next section).
Copies of one animation share their tagging
Sound designers often duplicate one animation to build a longer take: a two-step walk copied five times makes ten steps. You tag it once. Like REAPER's pooled MIDI items, the items of a project that play the same animation file are linked: they share one tagging (rules, thresholds, item options, preset, events and corrections).
One change, every copy. A threshold drag, a rule change, Analyse, loading, saving or updating a preset, an event correction: every linked item takes it at once, with its preview markers. Ctrl+Z puts every copy back together.
Each copy on its own place. The events are kept in animation time; each item places them on its own position, trim and play rate. A copy trimmed to the second step shows only the events of that step.
Commit and Cancel act on every linked item of the selected ones, even those not selected.
Edits in REAPER reach the copies too: drag, delete or rename a RAV marker of one copy, and the event changes on every copy; their markers follow at once and stay markers up to date.
The panel shows the link under the item's name: Linked · 4 other copies. Hover it for what is shared.
Make unique (the button on that line, or the action RAV: Tagging - Make tagging unique (selected items)) gives an item a tagging of its own, as it is now: from then on, a change on it no longer reaches the other copies, nor theirs this one. The line then reads Not linked, with a greyed Unique. Use it for a deliberate variation (a limp on one step). Duplicate a unique item and the duplicate is linked to it. To link an item back, undo the Make unique (Ctrl+Z): there is no "link again" button yet.
What makes copies: the same animation file (the same path, whatever its case or slashes), in the same project. Two different files never link, even with the same motion, and items in different project tabs never link. An item relinked to another file leaves its old copies and joins the new file's.
An item added later from the same file (no tagging of its own yet) is linked too: its panel shows the shared tagging at once, a change made on it changes the shared tagging, Commit writes its markers like the others', and Make unique keeps the shared tagging as its own.
Copies tagged differently by an earlier build are linked as well: the next change on one of them gives its tagging to every copy (Ctrl+Z undoes it).
Known limits
Smart Detection covers the legs (the feet's steps, lift-offs, slides and pivots, jumps and landings, taps and drags, heel taps, kicks, the knees on the ground), the arms (the hands' grabs, releases and pivots, slaps, claps and hands on the body, swings, punches and strikes, the elbows on the ground) and the body (its jolts and spins, sitting down, a fall onto the ground) today. Every other event is tagged with your own rules (see Tagging your own events), which you can save and share as presets. Punch was set on the sample punches (straight punches, jabs and crosses); Kick on two capoeira kicks and synthetic motions, so check their markers by ear.
Smart Detection: a touch shorter than about a tenth of a second is not a contact, a foot held still in the air reads as planted, and an animation exported in place (walking forward, or climbing) is read right only when a foot is planted for more than half of the clip. A slide or pivot gets one marker, at its start.
Smart Detection of pivots on a steep foot: on tiptoe, RAV reads the turn from the toe end, or, on a rig without one, from the toe bone's rotation (as long as the foot stands flat on its ball somewhere in the clip). On a rig with a toe end, a turn on the heel with the toes raised almost vertical (about 80°) is lost; with the toes lower, the heel and the toe still read it.
Smart Detection of the hands was set on a single tagged grab (a climb), so check its markers by ear. A hand that freezes in the air after a fast move (a long dance hit) reads as a grab.
Smart Detection of a turn that ends as the foot leaves the ground: under 30° (by default) it gets no pivot, even when you hear a scuff; raise the pivot sensitivity to catch it.
Smart Detection of hand pivots reads the hand's turn against how the hand rests over the whole clip: a hand that rests palm-down in one contact and flat on a wall in another may get no hand pivot.
Smart Detection of hand pivots has no tagged reference yet: it borrows the foot pivot's settings (how fast and how far the turn goes). Check its markers by ear.
Smart Detection of knees, elbows, sitting and falls has no tagged reference yet: it was set on the kneeling, crawling and falling sample clips (a knee rests about 10 cm above the floor, its joint's centre). No sample clip sits down or rests on its elbows, so check those markers by ear.
Smart Detection of claps and hands on the body has no tagged reference yet: no sample clip claps, and only a few touch the body. It reads the hand bone (the wrist) against the other wrist or the body's bones, so a touch on the far side of a thick body part (the belly, the back) can be missed. Check its markers by ear.
Roles are guessed from the bone names only (Mixamo and Unreal-style names). For another naming, set them once per skeleton in the Skeleton & roles window (click Roles).
In RAV view, hovering a joint in Skeleton shows its roles only once the Tagging view has been opened on an item with this skeleton.
The 3D view does not follow a take's play rate or a looped item, while the strip and the take markers do: on such items the pose and the markers can disagree.
Commit skips an item that has no rules left, and leaves the markers it wrote earlier: delete them in REAPER if you remove all of an item's rules.
A unique item can only be linked back to its copies by undoing its Make unique.
Two items that would write the same project marker at the same time (for example a character and its weapon on two tracks) share one marker.
Detection reads the bones' motion only: an animation built at runtime in the engine (IK, blend trees) must be baked into the exported file, as for viewing.
Saving and Sharing Projects
The viewer state (items, docked window position) is saved with your REAPER project and restored when you reopen it.
To send a session to a colleague, use File > Save project as... with Copy all media into project directory: the animation files are copied with the audio, and the project opens on another machine without re-importing anything.
When the animator delivers a new export, replace the file on disk (same name) and reopen the project: your items and sounds stay in place.
build.bat compiles the extension and installs it into REAPER (close REAPER first). Output: build/Release/reaper_animviewer.dll. Third-party dependencies are either vendored in extern/ or fetched and statically linked by CMake (see extern/VENDORED.md).
Build options, releases and the ReaPack workflow are documented in MAINTAINING.md.
