Manually made does not mean made without computers
Xin has stated that the film did not use artificial intelligence. The useful distinction is between manually directed computer animation and generated imagery, not between computers and no computers. Handmade is a colloquial description of doing the work, not proof that each image was drawn with a pen or that every frame was individually edited. His statement gives us a bounded account of his method. It is not a forensic examination of production files, and this guide should not pretend to have performed one.
Rough imagery is also insufficient evidence for identifying a tool. Unnatural movement, simple surfaces or a long-held shot may suggest a technical explanation to a viewer, but the result alone does not establish the process. When discussing a particular weakness, describing what is visible is usually more useful than immediately assigning an AI or non-AI label. This page is about completing and connecting production tasks. Why the film can remind people of AI, and how that association became a wider debate, is a separate question.
A character is more than one finished model
Xin describes a character-design stage involving fewer than ten figures and roughly three to four months of work. That is an approximate account of one stage, not a precise schedule for every character or for the whole animation. Its useful implication is that a recurring figure is not simply a picture made for a poster. When it appears in different scenes, viewers still need to recognise it. Decisions about detail therefore serve the continuing film, not only one attractive still image.
A simple example helps without requiring a software claim: a shape that reads clearly close up may be harder to distinguish at a distance, and a convincing front view may create different problems from another angle. This is an explanation of the task, not an allegation about an undocumented modelling incident in Niu Lai. Characters have to survive changing viewing conditions. Their recognisability, expressive movement and surface detail can therefore be assessed separately rather than compressed into the single question of whether a model looks attractive.
Why one distant landscape needed another approach
His concrete example is a distant mountain landscape with a rising sun. The intended scale proved inefficient for the available computer, so he revised the approach using his landscape-design experience. That establishes a constraint and a change of method. It does not establish which plugin he used, how many polygons he removed or how many days a render took. Even without those parameters, the example shows the gap between imagining an image and finding a practical way to make it available for the film.
That is more informative than saying a difficult shot was simply impossible. When a method changes, the important question is what the scene needs to do: establish space, create a sense of scale or prepare a change. These are ways to examine a scene, not invented explanations of the director's intention. For a small production, reaching a goal may require changing the approach rather than endlessly increasing detail. Viewers can still judge whether the final image works; the existence of a constraint does not make every constrained solution successful.
The limited resource is also attention
The technical workload was concentrated, and Xin describes having to keep the overall film and small details in mind rather than relying on separate specialists. Simultaneous responsibility does not mean literally doing every operation at the same second. It means returning to other tasks after solving one problem and remembering their connections. A scene that works in isolation may function differently between the shots before and after it. The challenge is not only the number of tasks, but repeatedly recovering their place within the whole.
A useful way to understand the workflow is through decisions, execution and checking the result in context, rather than only listing job titles. Deciding what an image needs, making the asset and checking its fit with movement, sound and timing are different activities. A change early in that chain may require another adjustment later. This explains why rework consumes time without proving that every pause in the film was caused by hardware or every movement problem by a lack of staff. Particular defects still need particular analysis.
Finished assets still need a finished sequence
Xin describes a final stage of combining completed material into an overall work. Characters, settings and sound existing separately do not mean a film already exists. For a viewer, when a sound arrives, how long an image remains and when the next piece of information appears all affect understanding. No particular editing application or export setting needs to be invented to explain that. Recognising the relationships between assets is enough to see why post-production is not simply a final button pressed after the important work is over.
A practical viewing distinction is between understanding what happened and feeling something about how it happened. The first concerns clear information; the second concerns rhythm and emotion. A sequence can explain an event without creating the intended tension or ease, while sound can add feeling and also reveal an awkward connection. Separating those layers produces a more specific discussion than saying the production is rough. This is an analytical method, not a request that viewers reinterpret every unsuccessful passage as a deliberate artistic choice.
What the available account does not establish
Claims circulate about the software, budget and number of computers, but the available account does not establish a complete inventory. This guide therefore supplies no specific animation package, graphics card, render duration or production price. Buying used hardware does not mean the entire film used one unchanged computer, and equipment failures do not explain every creative problem. Keeping these gaps visible separates a documented process from guesswork. It is not a suggestion that a secret or extraordinary technology must be hiding behind the result.
What this workflow guide can explain is how a small production faces connected decisions and how hardware and attention affect its methods. It cannot convert five years of labour into a quality score or offer a supposedly reproducible recipe built around an unverified software list. The useful way to respect the work is to describe it accurately and leave unknowns open. The useful way to assess the film is still to examine images, movement, sound and storytelling. Those two tasks complement each other without becoming the same judgment.