Fine hairs stretched between the towers of a model, or a light fuzz over a print that is otherwise clean. Usually worse on tall thin geometry with lots of travel between features, and much worse on some materials than others.
Look at the strings before changing anything
The strings themselves carry the diagnosis, and this is the step almost every guide skips.
Glassy, fine, straight hairs that snap cleanly are a pressure problem. The melt was under pressure, the nozzle travelled, and plastic came out.
Coarse, rough, slightly foamy strings, often on a print whose surface is also duller than it should be, and usually with audible ticks or hisses during printing, are a moisture problem. Water in the filament flashes to steam in the melt zone and blows plastic out of the orifice with far more force than residual pressure ever could. If you can hear it, stop and read popping and hissing instead, because no retraction setting fixes steam.
That distinction is why the community argument about this symptom never resolves. Both camps are right about different spools.
Why an idle nozzle leaks at all
There is a small reservoir of liquid polymer sitting above the orifice at all times. When the extruder pushes filament in, that reservoir does not respond instantly — the material compresses slightly, the filament itself acts as a long spring, and the barrel walls hold a pressure gradient. That stored pressure is what keeps flow going smoothly during a print.
The moment the toolhead stops printing and travels, that stored pressure has nowhere to go except out of the only opening. Retraction exists to remove it: pulling the filament back a short distance releases the compression before the travel move starts. It is a decompression, not a suction.
This also explains the machine dependency. On a direct drive the filament between the drive gear and the melt is a few centimetres long and barely compliant, so a small retraction does the job. On a bowden setup that column is much longer, so more of the retraction is absorbed by the tube and the filament stretching inside it before anything happens at the tip.
What to change, and what "better" looks like
Work in this order and print the same test each time — two thin towers spaced apart is the classic, and it takes minutes.
- Dry the spool if the material absorbs water. For PETG, polyamides, TPU and copolyesters this is the first move, not the last. It is free, and it is the fix for a machine that has "suddenly got worse" at unchanged settings.
- Drop the nozzle temperature. A hot melt is a thin melt, and thin melt oozes. Run a temperature tower and take the lowest step that still bonds properly. On PETG, moving from 250 to 235 °C often does more than any retraction change.
- Increase retraction distance in 0.2 mm steps on a direct drive, 1 mm steps on a bowden. Most direct drive machines land under 1 mm; bowden setups commonly need 4 to 6 mm.
- Raise retraction speed before raising distance further. A fast, short retraction relieves pressure better than a slow, long one, and it costs less time.
- Enable combing or travel avoidance, so the nozzle stays inside the part where any leak lands on infill nobody sees.
Success looks like fewer and thinner strings, not none. Some materials will always leave a wisp that brushes off.
Do not chase this with retraction alone
Pushing retraction distance up on a direct drive machine — past about 2 mm — starts pulling softened filament up out of the melt zone into the cooler section above it, where it swells and jams. That is heat creep, and it is a far more annoying failure than a few hairs. If a number that large seems necessary, the problem is temperature or moisture.
The other thing worth resisting is the reflex to blame the filament brand. Cheap PETG and expensive PETG both string when wet, and both stop when dry.
Materials that do this most
PETG and PCTG are the usual offenders, because the copolyester melt is both stringy by nature and thirsty. Polyamides string badly and are the most hygroscopic materials on this site. TPU strings because a soft filament transmits retraction poorly along its own length. PLA is the outlier in the other direction: a dry spool of PLA at a sensible temperature barely strings at all, which is why stringing on PLA is nearly always a moisture or temperature signal rather than a retraction one.