Neverenders is the feature. A24 is the producing entity. The script exists. The pre-viz is well underway. I am not going to spoil the story in this notebook and the company would not be thrilled if I tried, so what follows is a careful piece about process, not plot. There are three frame studies I can describe in prose because the prose does not give anything away that a viewer would not see in the first thirty seconds of the eventual cut. Everything else stays in the working archive.
Frame study one, the corridor
The first frame I locked in pre-viz is a corridor. A character walks the length of it, the camera holds steady at hip height, and the light moves across the wall as the character moves. It is not a complicated shot to describe. It is a difficult shot to plan because the corridor exists in three time periods inside the script. Same corridor. Different time. Different light. Different character age.
In a 2018 pipeline this would have been an art department conversation that took weeks. Three set dresses. Three lighting plots. Three wardrobe passes. A practical lighting test on stage with all three setups so the director could compare. In 2026 I generated the same corridor in Veo three times with the same architectural reference and different lighting prompts and different age states for the talent. The three frames sit side by side in my reference deck. The DP has seen them. The production designer has seen them. The script supervisor has seen them. Every department starts the conversation on the same page.
This is what I mean when I say AI pre-viz is load-bearing on Neverenders. It is not decoration. It is the document the production runs on.
Frame study two, the still face
The second frame is a still face. A close-up. The character holds. The camera holds. The cut does not arrive when you expect it. This shot is the spine of the picture and I will not say more about who or when, but I can say that the AI pre-viz of this shot taught me something I did not expect.
I generated the still face thirty times across two models. Veo for the character lock, Sora for the lens behavior. Of the thirty frames, three were the shot. Twenty-seven were variations that did not work because the eye line was wrong by a few degrees, or because the model leaned into a micro-expression that broke the stillness I needed, or because the breathing rhythm in the inference did not match what I wanted on screen. The three that worked taught me what the working version actually was, in a way that no storyboard would have. I cannot draw the difference between a still face that works and a still face that does not. I can see it. The model output gave me a way to look at it thirty times and pick.
In 2018 a shot like this would have been one storyboard panel labeled close up character holds. The actual interrogation of what the holding means would have happened on set, on the day, with a clock running. In 2026 the interrogation happens in pre-viz with no clock and no crew and no money on the meter. The shot is right by the time I walk on the set.
Frame study three, the wide
The third frame is a wide. The character is small in it. The world is large. The frame is built to make the viewer feel the scale of what the character is up against, without saying so. This is the kind of shot a director either earns or does not. A wide is risky. A wide costs screen time. A wide that does not earn its real estate is the most embarrassing shot in a picture because it sat there and asked the audience to look and the audience looked at nothing.
The pre-viz on this wide ran for two days. Sora for the camera language, because Sora knows what a 24mm anamorphic crane looks like and I needed it to know. The model produced about seventy passes of the wide. I picked four. Of the four, one is the shot. I have shown it to two people on the A24 team. They both held silent for a moment after the playback, which is the response I wanted, which is the only response a wide of that kind is allowed to earn.
I am telling you about it without showing you because the shot is the shot and the shot belongs in the picture and not in a notebook entry. But the fact that it exists in pre-viz at all, in 2026, with the quality it has, on a feature that has not started principal photography, is the part I want on the record. Neverenders is the first feature where the AI pre-viz is the document the production runs on from day one. Not as a side experiment. Not as a marketing piece. As the operational reference.
Why this matters
Pre-viz used to be a tool for communicating the director's intent to the crew. It still is. What is new is that pre-viz in 2026 is also a tool for clarifying the director's intent to themselves. The thirty face passes taught me what the face was. The seventy wide passes taught me what the wide was. The corridor in three time periods taught me what the corridor was. None of those clarifications would have happened in 2018 because none of them could afford to happen in 2018. The price was wrong.
The price is right now. The director is still the director. The shot is still the shot. The cut is still the cut. What is new is that there is no longer an excuse for not knowing exactly what the shot is for before the camera rolls.
Amos LeBlanc has been working on Neverenders since 2024. The campaign tracker for the broader 2026 positioning around AI directing is at amosleblanc.com/ai. The picture itself is in pre-viz. Principal photography will be announced when A24 is ready to announce it. Until then, this is the most I will say about the frames in the notebook.