What annealing actually does
Annealing means heating a printed part to a temperature close to, but below, the point where it turns soft, holding it there for a set time, then letting it cool slowly instead of yanking it out into room-temperature air. Nothing melts. What changes is the internal structure of the plastic.
During printing, each layer cools fast, and the polymer chains get frozen in whatever disordered arrangement they happened to be in when the nozzle moved on. That disorder carries two consequences: leftover internal stress from uneven cooling, and low crystallinity, meaning the chains never had time to line up into the tighter, more ordered regions that make a plastic mechanically stable at higher temperatures. Annealing gives the material a second chance. Held near its glass transition temperature for long enough, the chains gain enough mobility to relax that stress and settle into a more ordered, more crystalline structure. Slow cooling afterward locks that structure in instead of freezing in a new batch of stress.
It’s a known process borrowed from injection molding and metallurgy, not something 3D printing invented, but FDM parts are unusually good candidates for it because layer-by-layer deposition leaves so much stress behind in the first place.
What actually improves
The headline benefit is heat resistance. A standard PLA print starts to sag and deform once it gets into the 55 to 60°C range, which is why a PLA phone stand left on a car dashboard in summer is a bad idea. Properly annealed PLA can hold its shape well beyond that. CNC Kitchen’s published testing found annealed PLA maintaining structural integrity at temperatures that would have collapsed the untreated part, and Prusa’s own in-house testing measured close to a 70% improvement in heat deflection temperature after annealing at around 90°C for 30 minutes.
Mechanical strength moves too, though less dramatically than most people expect. In CNC Kitchen’s tests, tensile strength rose from about 63.5 MPa to 68.2 MPa, a gain of roughly 7.5%, and a more realistic loaded-part test showed about 16% more force tolerated before failure. Prusa measured around a 20% improvement in load-bearing capacity in its own trials. These are real numbers, but they’re not the "twice as strong" claims that circulate on forums.
What you give up
Nothing is free here. Two trade-offs show up consistently across independent testing.
First, impact toughness tends to suffer. Higher crystallinity makes a plastic stiffer and more heat-resistant, but stiffer plastics generally absorb shock less gracefully and crack instead of flexing. Some controlled tests, including CNC Kitchen’s drop and impact trials, found no measurable difference between annealed and untreated samples for PLA specifically, so this effect isn’t universal or huge for every material and every geometry. Still, treat an annealed part as more brittle by default, especially anything printed in a material known for high crystallinity gains, like nylon.
Second, and more disruptive in practice: the part shrinks, and it doesn’t shrink evenly. As the polymer densifies, dimensions in the print plane contract while the layer-stacking direction can actually grow. CNC Kitchen documented a 50mm dimension shrinking to 45mm in the XY plane on one test specimen, while the same part grew from 50mm to 55mm in Z. That’s an extreme case from a deliberately aggressive test, but even mild annealing runs typically leave a part a percent or two smaller in at least one direction, and asymmetric shrinkage means warping, not just uniform scaling. A bracket that was flat before annealing might not be flat after.
For background on other post-processing steps that don’t carry this kind of dimensional risk, see this post-processing overview.
How this plays out by material
PLA is the material people anneal most often, and for good reason: it has the lowest baseline heat resistance, so it has the most to gain. Guides and community tests generally land somewhere in the 55 to 65°C range for a first attempt, with more aggressive results reported closer to 90 to 100°C in controlled testing (Prusa ran its best results at 90°C; CNC Kitchen used 100°C for 45 minutes). Higher temperatures buy more heat resistance but also more shrinkage and warping risk, and PLA starts to visibly slump well above 70°C without support.
PETG behaves differently. Its glass transition sits higher than PLA’s to begin with, so it already tolerates more heat before annealing. Community guides commonly suggest a gentler starting point somewhere in the 65 to 75°C range, though Prusa’s own testing found the best balance of durability gain versus dimensional stability closer to 110°C, with the material staying visibly usable up to around 170°C before deformation became severe. The proportional shrinkage tends to be smaller than PLA’s for a given temperature step, but PETG parts are more prone to sagging under their own weight if unsupported during the process, so bracing or embedding matters more here.
Nylon is arguably where annealing does the most real mechanical work. Nylon crystallizes readily, and research on printed nylon composites has measured strength increases well over 50%, with some directions showing gains beyond 100% under compaction and heat treatment. This is also the material most associated with hot water bath annealing for smaller parts, since boiling water sits conveniently close to a useful treatment temperature for several nylon grades. Expect meaningful strength and stiffness gains, but also expect the brittleness trade-off to be more noticeable than with PLA.
ABS and ASA are the odd ones out. They already resist heat better than plain PLA, so annealing for extra heat deflection buys comparatively little. Prusa’s testing specifically flagged ABS and ASA as poor annealing candidates: the parts warped excessively at low treatment temperatures with negligible durability improvement to show for it. Where annealing does help ABS and ASA is stress relief, easing the internal tension that causes delayed warping or cracking weeks after a print comes off the bed, rather than pushing heat resistance further.
Practical methods
A kitchen oven is the most common approach, and it works, but household ovens are notoriously inaccurate. A dial set to 65°C might actually be running 10 to 15 degrees hotter or cooler, which is more than enough to either do nothing or collapse your part. A separate oven thermometer placed next to the print is worth the few dollars it costs, and it’s the difference between a repeatable process and a guessing game.
A hot water bath is the low-barrier option, mainly for small PLA parts. Bring water to the target temperature (not necessarily a full boil for PLA, since that can overshoot into deformation territory), submerge the part, hold it, then let the water cool with the part still in it rather than pulling it into cold air. It’s less precise than an oven, since you’re limited to water’s boiling point as a ceiling and stovetop temperature control is coarse, but it needs no special equipment.
Filament dryers and DIY annealing boxes sit in between. A filament dryer with a decent thermostat can hold a steady low temperature for hours, which suits PLA’s narrower safe window better than an oven that swings a few degrees around its setpoint. Some makers build small insulated boxes with a PID-controlled heater specifically for this, which is overkill for one part but worth it if you anneal regularly.
Keeping the part in shape
The shrinkage and warping problem has a few practical workarounds, none of them perfect.
Burying the part in fine sand or salt before heating is the most commonly recommended fix. The granular material acts as a loose mold, physically resisting deformation while the plastic is soft enough to move, without being rigid enough to trap the part or prevent even heat transfer. CNC Kitchen used this approach to reduce warping in more aggressive high-temperature tests.
Printing oversized to compensate for known shrinkage is the second approach, and it only works if you’ve already run a test piece and measured how much your specific material and geometry actually contracts. Shrinkage isn’t uniform across axes, so scaling a model up by a flat percentage is an approximation at best.
The third option is simply choosing which parts to anneal. Thick, simple, chunky geometry holds its shape far better than thin walls, long unsupported spans, or fine features, which will sag, bow, or curl before anything else happens. If a part needs annealing and needs to stay dimensionally precise, redesigning it to be simpler and thicker before you print it is often more effective than any trick applied after the fact.
A realistic expectation
Annealing at home is a fundamentally different process than annealing in an industrial oven with tight temperature control and forced-air circulation. Community results vary a lot between people running what looks like the same process, because small differences in oven accuracy, part geometry, and cooling rate change the outcome. Treat any temperature or shrinkage figure you read, including the ones above, as a starting point to test on a sacrificial piece, not a setting you can trust blind on a part that matters.
It’s also worth being honest about the ceiling here. Annealing a PLA bracket won’t turn it into a part rated for genuine high-temperature service, and it won’t replace a switch to a technical material like PETG, nylon, or a filled composite when the application actually demands sustained heat resistance or fatigue life. It closes part of the gap between a hobby material and a functional one. It doesn’t erase that gap.