Skip to content

Cleanup forecast ​

The cleanup forecast tells you how much work a take needs, before you start. A take is one recorded motion-capture performance, and a shoot day is the group of takes you captured on one day. Every take is rated right after it is diagnosed, so you can upload a whole shoot day and decide where to put people before anything is cleaned.

You see the forecast in two places on the Data screen, which is the screen that lists your projects and the takes in them. The row of a cleaned take carries the forecast as a chip. A take that has not been cleaned yet shows its predicted quality badge instead, because the badge and the forecast would use the same words for different things. Above the list, each shoot day gets a Cleanup forecast headline that reads like this: Shoot day · Aug 21 · 80% OK · 10% Review (predicted) · 10% Manual. While takes are still being diagnosed the headline says how much of the day it covers: covers 4 of 6 · 2 still diagnosing.

The four bands ​

A band is one of the four ratings the forecast can give a take. These are the four words you will see on a take row and in the shoot-day headline. Three of them are the same words as the quality badge, because they mean the same thing, and the forecast is a prediction made before anything is cleaned. None of them says Clean, so a forecast never reads like the take's state, such as Cleaned v2.

BandWhat it means
OKThe diagnosis found nothing. The take is kept as raw.
Review (predicted)An easy job. There is a small number of low-difficulty errors, typically short gap fills.
Significant effortMany error occurrences. The count alone makes it real work, even when each individual fix is routine. Budget time for it.
ManualMany errors and a lot of missing reference: long or one-sided gaps, swaps and merges at scale, props, large occluded or unlabelled stretches. Likely needs a person to verify the result. Post still attempts it automatically.

What drives the rating ​

The rating weighs the same two things a professional weighs when they open a file. The first is how many error occurrences the diagnosis found. The second is how difficult those kinds of error are to repair. Roughly, time ≈ count × difficulty.

  • Short gaps are easy.
  • Jitter and spikes are in the middle.
  • Swaps, merges, long or unanchored gaps, props and unlabelled stretches are the slow ones. Those are the "missing reference" that turns a big count into Manual.

So no findings at all is OK. A few easy ones is Review (predicted). A high count is Significant effort. A high count together with the difficult kinds is Manual.

How to use it ​

  • Let the OK and Review (predicted) takes run in the background. Auto is the run mode that cleans every take in To clean without asking you anything, so you barely look at these.
  • Put your review time on Manual, and expect to finish some of it by hand.
  • Use the shoot-day headline to plan a day rather than a take. A headline of 80% OK · 10% Review (predicted) · 10% Manual tells you how many people you need before anyone opens a file.

What the forecast is not ​

  • It is not a confidence score. Confidence says how cleanly Post can fix a take. The forecast says how much effort the take is. A take can be forecast Manual and still come back well cleaned.
  • It is not a guarantee. It is an estimate made from the diagnosis alone, before anything ran. The measured answer is the verdict on the Review screen, which is the screen you read once a run has finished.
  • It is not a gate. Nothing is refused for being Manual. Post still attempts it.
  • A take that has not been diagnosed yet, or was diagnosed before the forecast existed, carries no badge at all. Nothing is invented to fill the space.

What to do next ​

Sort a shoot day by its forecast. Send the OK and Review (predicted) takes through Auto. Open the Manual ones yourself in Assisted, which is the run mode where you choose the options per take. There the Cleanup Plan column will tell you what Post intends to do to each one.

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.