Does annealing actually make 3D printed parts stronger?

It makes them stiffer and more heat resistant, and usually less tough. Holding PLA above its glass transition lets the polymer crystallise, and crystallinity raises both modulus and the temperature at which the part softens. The same change costs impact resistance: a well-annealed PLA part is more likely to shatter than to bend when it is dropped or shock-loaded. So annealing is a sensible answer to a heat problem and a poor one to a breakage problem. If parts are cracking under load or impact rather than softening in heat, annealing makes the failure you already have worse.

How much does an annealed PLA part shrink?

Commonly one to three percent in X and Y, often with a small growth in Z, and the exact figure belongs to the part rather than to the material. Crystallisation raises density, so the part gets smaller; how that distributes depends on layer orientation, wall count, infill and where the mass sits. Two parts from the same spool in the same oven cycle can move by different amounts. That is why a scale factor copied from a forum does not transfer: print a coupon of the same geometry class, run it through the cycle you intend to use, and measure it before you commit a real part.

What temperature and how long should I anneal PLA?

Somewhere in the 80-100°C band, held for roughly 30-60 minutes once the part itself is up to temperature, then cooled slowly in the closed oven. Lower and longer is safer than hotter and shorter, because the failure mode above about 110°C is surface deformation rather than faster crystallisation. The number on the dial is not the number in the oven — domestic ovens routinely swing 15-25°C around a setpoint and toaster ovens cycle their elements against a single hot spot, so put a separate oven thermometer beside the part and set the dial to whatever makes the thermometer read correctly.

Can I anneal PETG, ABS or ASA for heat resistance?

Not for heat resistance, no. Those three are amorphous in practice, so there is no crystallisation available to them and nothing to raise the softening point. What a controlled heat cycle does do is relieve the internal stress frozen in during printing, which can stop a large flat ABS or ASA part from cracking at a corner days after it comes off the plate. Treat it as a stress-relief step with a modest dimensional cost, never as a way to upgrade the material. If PETG is softening in service, the fix is a different material rather than a different process.

How do I stop a part sagging while it anneals?

Support it everywhere, because above its glass transition the part is soft and is carrying its own weight. The standard method is to bury it in a fine, dry medium — table salt, dry sand or plaster — in an ovenproof container, so every surface is held and the whole mass heats and cools slowly. Failing that, lay flat parts on a flat tray rather than a wire rack, which prints its own grid into the underside. Thin walls, tall unsupported features and long overhangs are the worst cases and the least likely to survive a cycle unchanged.

When is annealing not worth doing at all?

When the part has to fit something, when it is loaded continuously, or when the real requirement is strength rather than heat. Anything with a mating feature loses its fit to the shrink, and no amount of care makes that shrink repeatable across geometries. Parts under constant load creep regardless of crystallinity, and parts that take impacts get worse rather than better. If the requirement is a stated heat-deflection figure on a drawing, that is a material and process specification, not an oven cycle — shops listed at /providers quote nylon, filled nylon and high-temperature polymers against the number you give them.

Why Annealing Warps 3D Printed Parts

The bracket printed cleanly. You read that annealing turns PLA into something that survives a hot car, put it in the oven at 100°C for an hour, and took out a part that is visibly smaller, bowed along its longest face, dished where it sat on the rack, and no longer lines up with the holes it was drilled to match. Nothing was wrong with the print. Annealing a 3D printed part moves it in three separate ways at once, only one of which is the effect you wanted, and the part in your hand tells you which of the three dominated.

This post contains affiliate links. If you purchase through these links, 3D Prototyping Hub may earn a small commission at no extra cost to you.

Read What Came Out of the Oven

Each of these points at a different mechanism, and they need different responses.

What you see	What happened	What to change
Uniformly smaller, shape intact	Crystallisation shrink — the intended effect	Nothing. Size the part for it next time
Bowed or twisted with no load on it	Frozen-in printing stress relaxing above Tg	Slower ramp, support the part, print hotter next time
Flattened or dished where it touched the tray	Soft part carrying its own weight	Bury it in a medium, or use a flat tray
Drooping overhangs, slumped thin walls	Same cause, worst-case geometry	Support medium is not optional on this part
Glossy, rounded edges	Oven overshot — surface approached melt	Thermometer in the oven, drop 10-15°C
Cracked at a corner or a layer line	Thermal gradient plus existing stress	Ramp up and cool down far more slowly
Holes no longer on their centres	Shrink is not isotropic	Anneal before drilling, reaming or fitting
Nothing changed but the colour	Amorphous material — nothing to crystallise	The material, not the process, is the limit

Annealing Is Crystallisation, and Crystals Take Up Less Room

PLA is semi-crystalline, but a printed PLA part is mostly not crystalline. Extruded plastic goes from melt to solid in seconds, which is far too fast for the polymer chains to fold themselves into ordered regions, so they freeze where they land. Hold that part above its glass transition — around 60°C — and below its melting range, and the chains become mobile enough to do what they could not do during printing. They crystallise.

That is the whole of the benefit. Crystalline regions resist softening, so the temperature at which the part loses stiffness climbs sharply; our hot-car piece records well-annealed PLA holding its shape in the 90-100°C region against roughly 60°C as printed. Stiffness rises with it.

It is also the whole of the problem. Ordered chains pack more tightly than disordered ones, so the part becomes denser and therefore smaller. There is no setting for this and no way to switch it off: the shrink is the mechanism you asked for, arriving as a dimension rather than as a property.

Three Movements, Arriving Together

Crystallisation shrink is the one people expect — typically one to three percent in X and Y, frequently with a slight growth in Z. It follows layer orientation and mass distribution, so it is not a uniform scale factor you can apply once and reuse.

Stress relaxation is the one that produces bows and twists. Every printed part has internal stress locked into it: each layer is laid hot onto a cooler layer below, contracts as it cools, and is held in tension by material that has already set. Below the glass transition that stress sits there indefinitely. Above it, the part is free to move toward the shape it wanted to be all along. Long flat parts, fast prints and cold open-frame machines carry the most of this, which is why an enclosure helps before you ever get to the oven.

Gravity is the one nobody plans for. Above the glass transition the part is not merely warm, it is soft — stiffness falls by orders of magnitude. For the whole dwell, your part is a slightly floppy object resting on whatever it was put on. A wire rack prints a grid into the underside. A tall feature leans. A thin wall slumps toward the horizontal.

Support It, or Accept the Shape It Chooses

The standard fix is to remove gravity from the equation by packing the part in a fine, dry medium — table salt, dry sand, or plaster — in an ovenproof container. Every surface is supported, and the thermal mass slows both the heat-up and the cool-down, which is the second half of the fix.

The rest of the cycle:

Ramp slowly. A part thrown into a preheated oven sees a steep gradient between its skin and its core, and that gradient is what cracks corners. Put it in cold, or bring the medium up with the part inside it.
Hold once the part is up to temperature, not once the dial says it is. A buried part lags the air by a long way.
Cool in the closed oven. Pulling a hot part onto a cold bench reintroduces exactly the kind of stress you just spent an hour relieving.
Distrust the dial. Domestic ovens commonly swing 15-25°C around a setpoint, and toaster ovens cycle a single element against a hot spot. A separate oven thermometer sitting next to the part is the only number worth acting on. Setting 100°C and achieving 125°C is how the glossy, rounded-edge failure in the table above happens.
Ventilate, and think about whether the oven you cook in is the oven you want to do this in. This is a judgement call rather than a specification, but it is one worth making deliberately.

Half the Filaments Have Nothing to Crystallise

This is the step that saves the most wasted evenings. Annealing does something useful only to a semi-crystalline polymer; for everything else the cycle buys stress relief and a dimensional change, with no gain in heat resistance at all.

Material	What a heat cycle actually does	Worth it?
PLA	Crystallises; large gain in softening temperature; 1-3% shrink; more brittle	Yes, on chunky parts with loose tolerances
PLA with carbon-fibre, wood or mineral fill	Fillers nucleate crystallisation and physically restrain movement — less distortion, smaller gain	Often the best-behaved candidate
PETG	Effectively amorphous; stress relief only, no change in softening point	No, if heat resistance is the goal
ABS / ASA	Amorphous; relieves printing stress, which can stop corner cracking	As stress relief, not as an upgrade
Polycarbonate	Amorphous; glass transition already around 140-150°C	No — drying matters far more
Nylon (PA6, PA12, filled grades)	Semi-crystalline; genuine gains, but the part must be dry first and will move again with humidity	Yes, with the moisture caveat
TPU	Nothing you want	No

If you are annealing PETG to make it survive a warm environment, stop. The material is at its ceiling already, and the hour in the oven costs dimensions for nothing.

The Dimensional Cost You Cannot Design Around Once

The reason annealing wrecks assemblies is that the shrink is not a scale factor. It is anisotropic, geometry-dependent and only loosely repeatable, so the 1.8% you measured on a test cube does not predict the bracket, and the bracket does not predict the housing.

Three rules follow, and they cover almost every annealing disaster:

Anneal before any operation that sets a dimension. Drill, ream, tap and press inserts after the cycle, never before. A heat-set insert that was perfect at printing is loose or crushed afterwards.
Never anneal a fitted assembly. Two parts that mate before the oven will not mate after it, because they do not move by the same amount.
If a dimension has to be right, qualify the cycle. Print a coupon of the same geometry class — same wall count, similar mass distribution — anneal it in the cycle you intend to use, measure it, and derive your compensation from that. The wider version of that diagnosis is in why 3D prints come out the wrong size, and the same warning about uniform scale factors applies here with more force.

The Cheaper Answer Is Usually a Different Material

Work backwards from the requirement. If the part needs to survive 80°C in a car window, or 100°C beside something warm, the repeatable route is a material whose glass transition is already above that number rather than a PLA part processed toward it.

ASA sits around 100°C as printed and brings UV stability that annealed PLA does not have at any crystallinity — the comparison across brands is in our ASA filament guide. Polycarbonate is in a different league again at roughly 140-150°C, and nylon is the choice where toughness has to come with the heat resistance; our nylon filament guide covers what the filled grades change.

All three ask more of the machine than PLA does, and that is the honest cost of this route. They want a hotend that reaches their temperature and still air around the part. QIDI's Plus5 is specified with a 370°C hotend and a 65°C heated chamber, which is the combination that makes these materials printable rather than merely purchasable, and Flashforge's enclosed CoreXY machines are a second manufacturer-direct route to the same thing. If you would rather keep the machine you have, start from the enclosure guide — an enclosure both enables the material and reduces the frozen-in stress that annealing later releases as a bow.

One caveat that catches people moving up from PLA: PC and nylon are hygroscopic enough that wet filament will undo the whole exercise, printing weak, bubbled parts out of a material you bought for strength. The dry box guide covers the storage side.

When Annealing Is Genuinely the Right Call

It is a good move in a narrow set of cases, and worth doing properly in those:

A chunky PLA part with loose tolerances that has to survive a warm environment, where you already have the part and a percent of shrink changes nothing.
Filled PLA, where the filler restrains the movement that makes this risky on neat PLA.
Stress relief on a large flat ABS or ASA part that keeps cracking at a corner after printing.
Nylon that will be machined afterwards, dried first, cut after.

And the cases where it is the wrong lever entirely, each of which has its own cause: parts that sag under a permanent load are creeping, which annealing does not fix; parts that break when dropped get measurably worse with crystallinity; and nothing in an oven addresses UV.

When the Number Has to Be Specified Rather Than Hoped For

There is a point where a kitchen oven stops being a process. If a drawing carries a heat-deflection figure, if a customer is waiting, or if the part has to behave the same across twenty units, you are asking for a specification rather than an improvement — and an uncontrolled cycle with a 20°C swing cannot deliver one.

That is the point to buy the property instead of processing toward it. SLS and MJF nylon, glass-filled grades and the genuinely high-temperature polymers arrive with published figures and no post-oven shrink to design around; high-temperature 3D printing services covers what those materials cost and why. Browse shops by location and process in the 3D printing directory, send the same file to two or three, and compare that against another weekend of oven cycles.

Hero photograph by Snapmaker 3D Printer on Unsplash.