Skip to content

FAQ ​

What does "Did not land" mean? ​

Post attempted that step and shipped nothing. It is not a crash and not a step that was skipped — it looked at its windows and put no value in the delivered file, so what it stood for is in your take as holes for an artist to finish. Running the take again will usually decline the same work the same way; the plan says so when that is the case. See Step states.

Why is the jitter cut the last step? ​

Because a cut takes frames out and has to rejoin the trajectory on both sides, and next to an unfilled hole there is nothing to rejoin to. So the cut runs after the fills of its own pass, once the surrounding data exists. That is also why there are two cut passes: filling a gap can expose a peak that was not visible in the raw. See Why the jitter cut comes after the fills.

Why is there still a hole in my take? ​

Either the frames were never recorded — the performer had left the capture volume, or the capture system dropped out — or a repair was rejected because it would have moved the marker faster than that marker can move. The After screen names which, with counts and where to find them. Re-cleaning will not fill the first kind: Post repairs occlusions, it does not invent capture. See Holes left open.

Where do I find the frames Post replaced? ​

On the After screen, in What the AI changed · per region. A card whose fill replaced measured frames says so on its face — Includes 111 measured frames replaced during filling (flickering signal) — and expands to list each stretch with its frames, the check that consumed it, and how far the fill ended up from the samples it replaced. See Fills.

My upload does nothing — the button is inactive ​

Create or select a project first. Uploads attach to a project, so a brand-new workspace has nowhere to put files until one exists, and the upload button stays inactive until then. Add a project from the + next to Projects in the left rail (or from the prompt in the upload dialog), select it, and upload again. See step 2 of the quick start.

I re-cleaned the same file and nothing changed ​

That's expected. Cleanup is deterministic per (raw, options, model) — the same raw file with the same options and the same model produces the same result every time. To get a different result, change the frame range or options, or upload a new raw file. Running Auto again over an unchanged take won't move it.

Why is a clean marked "unverified"? ​

Because Kapnetix won't show a fake score. When a result can't be measured, the Review screen labels it unverified rather than inventing a confidence number. This is deliberate: confidence is a real, measured figure on the After screen, and an unverifiable result is reported as such instead of dressed up. See confidence is fixability.

Can I convert formats on export? ​

Not yet — it is still to come. For now use Keep original: every supported format, TAK and VDF included, comes back in the format you uploaded. See Supported formats.

What happens to TAK and VDF files? ​

They come back as TAK and VDF. Both are proprietary containers with no public specification, so Kapnetix doesn't author one from scratch — it patches the cleaned animation into a byte-preserving copy of the file you uploaded, leaving your header, calibration, cameras and processing metadata exactly as they were. Validated by opening the results in MotionBuilder, Maya and Shōgun.

A solved-skeleton FBX is the one exception: it comes back as a fresh build rather than a patch, because a solve goes stale once markers move. See Supported formats.

Does cleanup ever overwrite my raw capture? ​

Never. Raw capture is immutable. Every cleanup lands as a new named version — Clean v1, Clean v2, and so on — and the version history is never collapsed. You can compare any version against the raw and revert individual regions. See Raw is immutable.

What does a confidence score actually mean? ​

How cleanly Kapnetix can fix the take or region — not how broken it is. You see it twice: Before it's predicted (plan your review), After it's measured (verify the result). It is never an error meter. See core concepts.

How long does diagnosis take? ​

From seconds to minutes for most takes. Long, multi-performer takes take longer — diagnosis and cleanup each get up to two hours, so a large file runs to the end instead of being cut off. Watch the take row on the Data screen: it shows Diagnosing… and then resolves to a status and a Cleanup forecast.

What this page is not ​

A reference. It answers the questions people actually ask us. For how the cleanup itself works, read how Post cleans a take; for the release history, see what's new.

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.