The client's model looks fine on its own—why does it render with tilted buildings and wrong proportions?

The model sent by the client looks fine on its own, but when rendered, the building is tilted and the scale is wrong. In most cases, this is not a renderer problem—it's that unit scale and coordinate origins were never aligned during the modeling stage. Based on common delivery experience in 2026, before taking over, run four checks first: units and scale, coordinates and origins, materials and naming, assets and paths. Only after all four pass is it even worth discussing whether the model can be used directly. The judgment can be boiled down to one line: it's only usable if it can be aligned, can be modified, and can produce a final image.

Why 'opens normally' is still a long way from 'can produce a final image'

On the modeler's own computer, the model is a complete set of relationships: layers, components, and material spheres all point to local files. Once the file moves to another machine, those relationships can break. A renderer is pickier than the browse mode of modeling software: it needs to resolve units, normal directions, texture color spaces, and material channels, and then unify them into one camera and lighting setup. So the common state when taking over is 'the structure is right, but it needs fixing.'

Another more subtle difference is the purpose of modeling. Client-side modeling is often for design exploration, area statistics, or construction coordination, and how dense the lines are or how much detail is kept or dropped is set by those goals. A visualization image needs the picture to hold up: scale relationships, material zoning, and opening edge details must withstand the camera moving in close. The two passing standards are not in the same place. This also explains the client's line, 'I already gave you the model, so why does it still take so long?'—the time isn't spent on rendering; it's spent on alignment before rendering.

• Browse-level model: can be viewed and used to measure area, but does not necessarily hold up to rendering;
• Render-level model: unified units, correct normals, clear material zoning, complete texture paths;
• Image-level requirements: under a close-up viewpoint, edges remain intact, reflections have a source, and entourage proportions hold up.

When buildings look tilted or proportions are wrong, I usually run the four checks first

Rendering rework looks all over the place, but at the root it usually comes down to four chains: geometry, coordinates, materials, and assets. The first three determine whether the image can hold up; the fourth determines whether the delivery can leave your own machine. Only after all four checks pass do you decide which path this version of the model should take—not render a version first and see what happens.

1. Check units and scale: measure a known dimension, such as a door opening, floor height, or column spacing, and verify it against the drawings; confirm whether the unit is meters or millimeters and whether any scaling remnants remain;
2. Check coordinates and origins: see whether the origin falls on the project base point, whether models from different disciplines share the same coordinate system, and whether anything has been casually moved;
3. Check materials and naming: see whether faces are zoned by material, whether the entire building shares one material sphere, or whether the same surface uses seven or eight similar grays;
4. Check assets and paths: determine whether textures are packed/embedded or use local absolute paths, whether anything has been renamed or lost, and whether proxies and reference files are all included;
5. Judge one of three tiers: directly usable, usable after local patching, or recommended for rebuilding based on confirmed drawings. The judgment should state the reason—don't just reply that the model is no good.

The order cannot be reversed. The first and second checks are hard thresholds; if they are wrong, the perspective and proportions of the entire image will be distorted, and no amount of later camera adjustment can save it. The third check affects whether, when the client changes the design, materials can be replaced in batches by material group. The fourth check often blows up when the file is handed to an external partner or when rendering on a different machine. In experience terms, the four checks themselves usually take 0.5 to 2 working days, with project scale and layer chaos determining whether it falls toward the upper or lower end.

• Directly usable: units and origins are correct, material zoning is clear, textures are complete; just add cameras, lighting, and entourage;
• Local patching: the main structure is reliable, but waste lines need cleaning, materials need re-zoning, and layers need organizing; typically 1 to 3 working days;
• Recommended rebuild: key structures are missing, overall proportions are off, or there is no confirmed drawing set to check against; the time to keep fixing approaches a redo.

The half day that often shows up in delivery

In projects, this kind of schedule comes up often: the client sends the model on Friday afternoon, a review needs images by Monday morning, and the budget only covers one round of changes. The approach is not to reply 'yes' when the file arrives, but to run the four checks that evening, list the four issues—units, origins, material zoning, and texture paths—and send them back so the client can confirm which items are handled by whom and against which drawing version. The cost is that the review images are delayed by half a day to a day. But if you skip this step, the common outcome is discovering halfway through rendering that the entire building's proportions are wrong, or only after delivering the final image realizing that reflective surfaces use blank material spheres. The rework then goes from half a day to two or three days, and cameras, lighting, and entourage all have to be redone. For this kind of handover, experience range usually reserves 0.5 to 1 working day for a health check.

• Looking only at file size and not layer structure: the file is huge, but it's actually all fragmented lines and hidden objects;
• Judging whether it can render from the preview in the modeling software, ignoring how different renderers parse material channels;
• Components or proxies from other projects are mixed into the client model, so unrelated buildings appear when rendering;
• Textures come from casually downloaded assets, and their source and licensing terms were not checked before delivery;
• Doors, windows, and curtain walls built by the client to area-based standards; once the viewpoint pushes in, the thickness and edge details don't hold up.

Direct rendering, local patching, or rebuilding: how do the three paths compare?

Which path to take depends mainly on how close the viewpoint is to the main subject. Medium and distant views focus on overall proportions, massing relationships, and entourage density; as long as the client model has correct coordinates and correct massing, it can often be used. Eye-level close views and detail shots focus on edge details, thickness, material joints, and reflection details—exactly the parts most easily omitted during design exploration. Using a client model for close views has a noticeably higher probability of rework.

• Direct rendering: suitable when the model itself was built to render standards, there are confirmed drawings, and viewpoints are mainly medium to distant; typical experience range is 1 to 3 working days; the trade-off is low flexibility—if the client changes the design on site, the whole version may need to be redone.
• Local patching: suitable when the main structure is reliable and what's missing is layers, materials, and details; typical experience range is 2 to 5 working days; the trade-off is that patching may uncover new structural problems, requiring one additional confirmation, so schedule buffer should be reserved.
• Rebuilding: suitable when the model is unreliable, close views are required, or dimensions must be checked against permit/approval submission standards; typical experience range is 3 to 10 working days; higher upfront investment, but more stable when changing viewpoints or revising the design later.

It's also important to distinguish 'saving modeling time' from 'saving the total cycle.' Taking over a client model saves the labor of modeling from scratch, but the four checks, patching, and drawing verification shift into the pre-rendering preparation stage. If there are many viewpoints and the number of revision rounds was not agreed well, the total cycle is not necessarily shorter than rebuilding.

When it's suitable to take on the model, and when not to force it

Situations suitable for directly taking on a client model: the model has corresponding confirmed drawings, the main structure and floor count are sound, viewpoints are mainly aerial, semi-aerial, or distant, and the review milestone leaves half a day to a day for handover verification. For such projects, experience shows the modeling labor saved is commonly 20% to 40% (experience range, depending on model quality), and the saved time can go into lighting and entourage.

• The model is only a conceptual massing with no dimensional basis, used solely for massing exploration;
• The client explicitly requires dimensions to be checked against permit/approval submission standards; using a model of unknown origin with no drawings to verify against carries higher risk;
• The viewpoint falls on interior close-ups, product detail shots, or locations requiring precise edge details and precise reflections;
• The client requires images the same day or early the next morning, leaving no time for the four checks and confirmation.

A boundary sentence can be written directly into the communication record: whether a client model can be taken on depends on viewpoint distance and whether the drawings can be checked against, not on whether the file can open. The former determines whether it needs fixing; the latter only determines whether you can see it.

Common questions

If the model from the client is a SketchUp file, does it need to be converted to another format first?

Most renderers can directly import common 3D formats, but importing does not mean usable. After conversion, first check whether units, group hierarchy, and material naming have been broken apart, then run the four checks again, and only after that schedule rendering.

If the client provides the model, how much cheaper will the visualization production quote be?

What's saved is modeling labor, commonly around 20% to 40% (experience range). If the model requires large-scale patching, that saving is offset, and the quote depends more on the number of viewpoints, number of final images, and revision rounds.

If textures are missing from the client model, does recreating materials count as extra work?

Yes. Missing textures count as incomplete assets, and are generally billed as material rebuilding, commonly adding 0.5 to 2 working days (experience range), depending on the number of materials and whether reference images are available for matching.

If the client changes the design after taking over, does that count as model changes or rendering changes?

It depends on whether the change is geometry or the image. Moving walls, replacing doors and windows, or adding floors counts as model changes and triggers re-rendering and camera fine-tuning; changing only material colors, entourage, and time of day usually counts as a rendering revision.

The client model has a very high polygon count; will direct rendering cause problems?

A high polygon count alone does not necessarily crash things; more commonly, waste lines and hidden objects are eating resources. First clean unused layers and merge duplicate components, then decide based on the viewpoint whether to reduce polygons. Don't simplify the whole model right away.

Next time you receive a client model, don't reply 'yes' right away. Run the four checks and one judgment, write the checklist and tier decision clearly, and then quote and schedule. It is suitable for projects with viewpoints mainly at medium to distant range and with confirmed drawings to check against. If same-day image delivery is required, or if the viewpoints fall on close-up edge details and permit/approval dimension verification, taking it on directly is actually more expensive. Based on common delivery practice in 2026, this kind of handover usually reserves half a day to a day for a model health check.