Skip to content

Supported formats ​

Kapnetix Post reads the common motion-capture interchange formats and writes back standard, editable files that drop straight into your pipeline. A take is one recorded performance, held in one of these files. The format and the tags are detected when you upload, so you do not pick the type by hand.

FormatSourceReadExport
FBXAutodesk (Maya, MotionBuilder)✓Keep original
C3DMarker-cloud standard✓Keep original
BVHSkeleton / hierarchy✓Keep original
TRCMarker trajectory (Motive / OpenSim)✓Keep original
TAKOptiTrack Motive✓Keep original
VDFVicon Shōgun✓Keep original

Every supported format round-trips in its own format. In the Export dialog the option reads Keep original (C3D), or Keep original (VDF), and so on for each format.

Below the format cards, Convert to another format is shown in a dashed box marked NEXT: it reads Converting between formats is not available yet. Use Keep original. Until it lands, use Keep original.

The rule behind all of it ​

Where Post can, it writes your clean by patching your own file instead of building a new one from what it loaded. Everything the file carried that Post does not need to touch is still there afterwards, because it was never rebuilt.

Inside the marker data, Post changes captured frames only where a region card says so. A fill writes the frames that were missing. A repair may also replace captured frames it judged unreliable, such as a spike or the flicker between drop-outs. A spike that Post could not repair is left as it was recorded, even inside a gap Post filled, and its region card says so. See Spikes the diagnosis named far over the limit. And up to two frames on each side of a gap may move, each by at most a short distance, so a fill joins smoothly. Every region card says what it changed, and a card where frames beside a gap moved also says how many and the farthest distance in centimetres. See Captured frames beside a gap. Where the usual repair of a gap could not be completed, Post fills the gap as a best attempt. Its region card starts Filled this 9-frame gap as a best attempt, says why the usual repair could not be completed, names the model as Best-attempt fill and asks you to check those frames by eye before you export. See Best-attempt fills.

On the Review screen, the line about moved captured frames sits on its own under what the card says it changed. Sometimes a short stretch of captured frames sits inside a gap, and the diagnosis did not flag it as wrong. Post then keeps that stretch close to where it was recorded and joins the fill to it, and the card says so: Kapnetix kept 7 captured frames inside the gap within 0.4 cm of where they were recorded and joined the fill to them. The panel above the region cards adds one sentence for each of these, and only when the count is not zero:

  • The usual repair could not fill 1 gap, so Kapnetix filled its 5 frames as a best attempt. Check it by eye before you export.
  • Kapnetix left 4 frames in 1 gap open because any fill there would move the marker faster than a real body part can move. Those frames ship with no value, and you finish them by hand.
  • Inside a gap, Kapnetix kept a short stretch of captured frames (4 frames) as it was recorded and joined the fill to it.

What that preserves depends on the format, and the sections below take each format in turn. One thing does not change in any format: the solved skeleton inside the file is the one you uploaded. Post is a marker-space cleaner, which means it corrects the marker cloud rather than the rig. A solve made from the uncleaned markers no longer matches them once they are corrected, so the deliverable needs a re-solve in your own tool for the corrections to reach the rig.

C3D ​

A cleaned C3D is written by patching your uploaded file. The corrected markers go into the channels their labels already occupy. Channel order, every parameter group, the header and the analog block are your file's own, because none of it is rebuilt. A pipeline that addresses markers by channel index gets back the channels it sent.

A save that changes nothing returns a file byte-identical to the one you uploaded. Template channels that no cleaned performer claims are left untouched.

One limit is worth knowing: positive residual magnitudes and camera masks do not survive a C3D write, patched or not. The point block is re-encoded and only the "invalid sample" marker is carried through, which is what a gap needs and all it needs.

If your original is not available at export time, the C3D is rebuilt from the markers instead, and the export note says so:

Rebuilt from the markers alone — your channel order and parameter groups are not preserved.

TAK and VDF ​

TAK (OptiTrack Motive) and VDF (Vicon Shōgun) come back as TAK and VDF. Both are proprietary containers with no public specification, so Post does not author one from scratch. It patches the cleaned marker animation into a byte-for-byte copy of the file you uploaded. Your header, calibration, cameras and processing metadata survive untouched, and a save that changes nothing returns the file you gave us.

We validated this by opening the exports in MotionBuilder, Maya and Shōgun.

Because the container is copied, the solved skeleton inside it is copied too, and it is unchanged. The export dialog says so:

Cleanup corrects the marker data; the solved skeleton in this file is unchanged. Re-solve in Shōgun to see corrections on the rig.

For a TAK the same sentence reads Re-solve in Motive.

These two formats are the one case where the original file is required. Without it there is nothing to patch, so the export falls back to FBX and says why in the export note. See the export note.

FBX ​

An FBX is patched the same way when your uploaded file is available. Only the marker translation curves are replaced, along with the occlusion channels where those are present, so your rig, props, scene graph, node names and properties survive.

There are two limits.

  • A rig-bearing, solved skeleton is not patched. A solve goes stale the moment markers move, and re-solving belongs in your tool, so a rotation-bearing skeleton is refused rather than written over. That take comes back as a fresh build with its animation intact.

  • Without your file, an FBX is rebuilt from the markers, and the export note says what that costs:

    Rebuilt from the markers alone — your rig, props and scene graph are not in this file.

BVH carries a solved skeleton and no optical marker cloud, so there is no marker cleanup to plan for it. The Cleanup Plan column, which lists the steps Post will take on a take, says Nothing to plan for those takes.

Performer names on marker labels ​

A performer is one person in the capture volume. When two performers wear the same marker set, both of them have a marker with the same name, for example BackLeft. MotionBuilder cannot hold two markers with the same name, so it renames one of them, and it does not always rename the same one. A retarget set up on one take then breaks on the next take.

Post keeps marker names exactly as the file has them, unless you ask it to add the performer's name. Adding it prevents the renaming: Post puts the performer's name in front of every marker name in the cleaned C3D and FBX. The names come from the file when it has them, for example Red:BackLeft and Blue:BackLeft for a file that labels its performers Red and Blue. When the file has none, the markers are named A1, A2 and so on, for example A1:BackLeft and A2:BackLeft. The number follows the order the performers appear in the file, so the names are identical across the takes of a shoot.

The choice is Marker names in the output. It sits in Assisted on the Clean screen and applies to the whole run. It has two options:

  • Keep as in the file is the default. Marker names stay exactly as the file has them, and nothing is added. Auto always keeps the names as in the file.
  • Add the performer name puts the performer's name in front of every marker name, as described above. When the diagnosis shows two or more performers sharing marker names with no performer name in front, this option is tagged Recommended, with a sentence saying why. It is still not selected for you.

Post never removes performer names that are already in your file, so a file that names its performers keeps those names whichever option you pick. Your choice is shown again when you export.

After a clean with Add the performer name, the Review screen says where each name came from, under the take name. For a file that names its performers it reads, for example, The performer names Red and Blue come from the file. For a file with one named performer it reads The performer name Red comes from the file. When Post wrote the names itself it reads The app named A1 and A2 because the file has no names for these performers, or The app named A2 because the file has no name for that performer when only one performer had no name. Cleans made before this was recorded show nothing there.

TAK and VDF files are not affected, because those formats carry performer names of their own.

The export note ​

The export note is the sentence the Export dialog shows when the delivered file is not exactly what you asked for. Post never delivers a substituted file without saying so. There are three sentences you may see, and silence when nothing was substituted.

The format changed. The note gives what you asked for, what was written, and why, in the cleaner's own words:

You asked for VDF; this Clean version was written as FBX. VDF needs the original file as an export template; none available.

The format changed and nobody said why. The served file's own extension disagrees with what you uploaded, and no explanation came with it:

This file was written as FBX, not the VDF you uploaded. The cleaning worker did not say why.

The format is right but your file was not used as the template. This is a fidelity note rather than a format change, and the two kinds never stack:

Rebuilt from the markers alone — your rig, props and scene graph are not in this file.

Rebuilt from the markers alone — your channel order and parameter groups are not preserved.

Either of the "rebuilt" lines may be followed by the reason the template was not used, for example No source file available to patch.

The same notes appear on the take's version history, so a Clean version carries its own export story rather than only appearing in the dialog you happened to have open.

What Post does not promise ​

  • No format conversion on export yet. C3D → BVH and the other conversions are still to come.
  • No solving. See the re-solve note above. This is a division of labour, not a gap.
  • Joint layers are re-derived, not copied, where they are written at all. Marker data is exact, and bone animation is faithful rather than bit-identical.
  • Nothing is written over your raw capture. Your raw capture is the file you uploaded, and every clean is a new Clean version beside it.

Capture systems and frame rate ​

Post works with takes from Vicon (including Shōgun .vdf), OptiTrack (including Motive .tak), and markerless capture.

Frame rate, duration, marker count, actor count and capture system are read from the file itself, and they are shown on the take's stats panel after diagnosis. Nothing is assumed.

What to do next ​

Upload one take in the format your pipeline actually uses, clean it, export it with Keep original, and open the result in your own tool before you commit a whole shoot day.

Ask us ​

Something looks wrong on a take? Write to support@kapnetix.ai and send the take id, or the take's name and the project it sits in.