Which way up to print it, and where the supports go

Within a layer, the plastic is continuous and strong. Between layers, it is a thermal weld formed when hot plastic was laid on plastic that had already begun to cool. That weld is real and it is not as strong as the material.

So a printed part has a grain, in the same sense that wood does. Load it across the layers and it behaves like plastic; load it so it is trying to pull the layers apart, and it behaves like a stack of things that were stuck together. A hook printed flat on the bed is strong; the same hook printed standing up will break at the bend under a load the first one shrugs off.

Strong direction

Within a layer

Weak direction

Across layers, in tension

So orient for

Layers perpendicular to the pull

• 1. Where is the load, and which way does it pull? — If the part has a job, this wins over everything else. Work out where it will break and rotate the model so the layers run across that pull rather than along it. A part that fails is a part reprinted, which costs more time than any surface finish saves.
• 2. Which surface has to look right? — The face pointing up is the best-looking one. The face on the bed is flat and shiny but takes the bed’s texture. Any face touched by support will be rougher than both. Decide which face the person looking at this will see.
• 3. What can be printed without support at all? — Below roughly forty-five degrees from vertical, an overhang is mostly supported by the layer beneath it. Above that, it increasingly is not. Rotating a part by a few degrees frequently takes every overhang inside that limit and removes the support question entirely.
• 4. How tall is it, and how narrow at the base? — Tall and thin prints wobble, and the head knocking a tall part over is a common failure. Lying it down is often the whole fix. Height also drives time far more than volume does, because every layer costs the same overhead however little is on it.
• 5. Would splitting it be better than any orientation? — Two parts each printed in their ideal orientation and then joined is frequently stronger, faster and better-looking than one part printed in a compromise. A printed alignment pin and a suitable adhesive makes a joint that is not the weak point.

• Support is scaffolding, and it leaves a scar — Every surface a support touches comes out rougher. So the real question is never "how do I support this" but "how do I need less of it", and the answer is usually rotation.
• Bridges are not overhangs — A flat span between two points prints fine over surprising distances, because the plastic is stretched between two anchors rather than hanging off one. A hole in a vertical wall does not need support; the top of it is a bridge. It will come out slightly droopy and it will be fine.
• Support on the build plate only, where it works — Support that grows from the bed is easy to remove. Support that grows from the model itself always marks the model. If a setting will restrict support to the plate without leaving something unsupported, use it.
• Change the model instead, if it is yours — A chamfer under an overhang, a teardrop-shaped hole instead of a round one, a forty-five degree bevel where there was a flat shelf — these are the standard ways of designing a part that needs no support at all, and they take a minute in the model.
• Interface layers are worth the setting — A denser layer of support directly under the part, with a slightly larger gap, gives a much cleaner surface and much easier removal than support that runs straight into the model.

None of this is specific to a slicer, and every slicer names these things differently. What is worth knowing is that the numbers a slicer offers — overhang threshold, support density, the gap between support and part — are trade-offs between removability and surface quality, and moving them is normal rather than advanced.