The Episode We Deleted

We deleted it. Not shelved, not cut down, not held back for a quiet week. Deleted, project file and all, and this is the account of that, written mostly because we have never seen anyone else write one.

A forest episode. Late afternoon into evening, a path, a clearing, a small stone shelter with a lamp in the window that comes on about eleven minutes in. Nothing happens, which is true of everything we make.

On paper it was the strongest brief we had had in months. The colour work is still the best we have done — there is a passage where the light goes from gold to green to blue over about four minutes that we have never matched since. The trees have weight. The sound bed is excellent. If you had shown us a still frame from it we would have said yes, that, please, more of that.

And watching it was like reading a beautifully typeset page in a language you do not speak.

Three wrong theories

We spent nearly two weeks being confidently wrong about the reason, which is worth recording, because each theory sounded correct at the time and produced work that changed nothing.

1. 1

   It's too slow

   The obvious one, and the first thing anybody says about work that isn't landing. We tightened the first eight minutes, brought the lamp forward, added a bird crossing the clearing. The episode got faster and stayed dead. In hindsight this theory was never plausible — our other episodes are slower, and people fall asleep in them happily.

2. 2

   It's too cold

   The palette theory. Forest evening means green and blue, and we build warm rooms, so perhaps the whole thing was fighting the channel. We regraded the middle section warmer, pushed more of the lamp's light into the clearing. It looked lovely and it changed nothing at all. This is when the first doubts started, because a regrade that does nothing usually means you have misdiagnosed the level you are working at.

3. 3

   It needs an animal

   The lesson we had learned on the lantern workshop, misapplied. We had a cat rescue that episode — the whole story is in the lantern maker's workshop — so we tried a fox at the treeline, twice, at two different distances. Both versions were worse. The fox arrived, the frame acknowledged it, and then it left and the episode went back to being what it was. It felt like a visitor to a place that was not expecting anyone.

Three fixes, ten days, no movement. The tell, looking back, is that each fix made the episode better by every local measure and no different by the only one that mattered.

The diagnosis, as close as we got

Somebody said, on about the eleventh day, that the shelter looked like it had been built for the episode. And that was it — or that was the closest anyone got to it.

Nobody lived there. Not in the sense of an empty chair or a cup left half-drunk; we had all of that, we know that trick, the shelter had a jacket on a hook and a pair of boots by the door. The problem was underneath the props. The building sat on the ground the way a model sits on a table. The path went past it rather than to it. The lamp came on because eleven minutes had elapsed, not because anyone had come home.

Our rooms usually have a history that we have not shown you and mostly have not even decided. This one had a set dressing list.

The difference between a room and a set is not what is in it. It's whether the space existed before the shot did.

We still cannot tell you how to check for that in advance, and we have looked. It is not the same as the used-object rule, which we know how to apply and did apply here. Something about a place has to be settled before it is drawn, and we do not have a procedure for it.

The room test

The one useful thing that came out of those six weeks is a question we now ask on every episode, usually at the first full playback.

Could you walk into this and sit down?

Not: is it beautiful, is it finished, is it on brief. Just — if the frame were a doorway, would you step through it and stay. It is embarrassingly unrigorous and it is the most reliable note we have. People answer it instantly and they answer it honestly, because it is not a question about the work, it is a question about themselves.

Why we did not just publish it

We could have. It was, by any external standard, a finished episode of the kind we make. It would have gone up, it would have done reasonable numbers, most people would have found nothing wrong with it. Several people would have liked it.

The argument against was not artistic integrity, which is a phrase we distrust. It was narrower than that: if we ourselves would not put it on at eleven at night, then publishing it means asking someone else to spend an evening on something we would not spend an evening on. We did not want to find out whether we could get away with it.

There is also a version of this decision that is just pride, and we cannot entirely rule that out.

What it cost

Honest accounting, because the number is the only part of this that is not a matter of opinion.

Six weeks of production, plus ten days of failed rescue. Roughly ninety hours of render time on the final pass alone, which is a real electricity bill. One publication slot missed, filled with a short re-cut of an older episode that we are not proud of. And a fortnight afterwards where nobody wanted to start anything, because starting something is easier when you have not just watched six weeks go into a folder and out of existence.

We kept two things: the light transition, which we have since stolen from ourselves twice, and the sound bed, which is now under a different episode entirely and nobody has noticed.

We do not fully know why it failed

This is where a piece like this is supposed to turn — where the failure becomes a lesson, the lesson becomes a principle, and the principle makes the six weeks worth it in retrospect.

We have the room test, and the room test is genuinely useful, and it is not worth six weeks. It diagnoses; it does not explain. We still do not know why that particular forest, built by the same people using the same methods as a dozen rooms that worked, came out as a place nobody had ever been in. We have theories. We had theories at the time, and three of them were wrong with total confidence.

The nearest thing we have to a conclusion is that some work does not come together and the reason is not always available. That is an unsatisfying sentence and we have tried several times to write a better one.

If it helps, the opposite story is on the site too: the lantern maker's workshop is an episode that was dead in almost the same way and came back for reasons we could name. We would love to tell you what the difference was.

The evening is warm here.