Fix TPU Stringing: Why More Retraction Makes It Worse

Prusa’s own knowledge base lists stringing, oozing, and filament tangling into the extruder gears as standard cons of any flexible material, not a defect you introduce by getting a setting wrong. That framing matters here, because it means the goal on TPU is not zero strings the way it might be on PLA. It is keeping stringing survivable and easy to clean up, while avoiding the setting changes that turn manageable stringing into a jam. For the rest of the profile, temperature, bed, and speed beyond retraction, see TPU Print Settings.

Why TPU strings even when the numbers look right

PLA and PETG are rigid enough that a retraction move does what the name says: the drive gear pulls a length of solid filament back, away from the melt zone, and that filament arrives at the nozzle tip a moment later ready to push forward again. TPU is an elastomer. Squeeze it in the drive gear and part of that pull turns into stretch instead of movement, the same way tugging on a rubber band takes up slack before anything at the far end actually moves. Some of what should be a clean retraction becomes stored elastic energy that has nowhere useful to go.

Prusa’s flexible materials guide gets at this indirectly with an instruction that would look strange on any other filament: loosen the idler so the gears only push gently, because pressing too hard is what tangles the filament into the gears in the first place. Bambu Lab’s TPU guide describes the same failure from the other side: soft grades get flattened by the extruder wheel instead of feeding cleanly, and a flattened section jams rather than retracting. Different vocabulary, same mechanism. The filament is absorbing force instead of transmitting it.

Why more retraction makes it worse, not better

The instinct carries over directly from PLA and PETG: strings showed up, so add retraction distance until they stop. On TPU that instinct is backward, and Prusa says so almost bluntly. Its guidance for flexible filaments is to turn retraction off completely first, then raise it in small steps while watching what actually happens, rather than starting from a generous number and hoping. Raising retraction on a filament that already compresses instead of pulling back cleanly does not relieve more pressure. It just gives the filament more room to buckle sideways and wedge itself against the gear or the tube wall, which shows up as grinding, a skipped step, or a full jam, not as fewer strings.

There is a real contradiction worth naming here rather than smoothing over. OrcaSlicer’s own documentation on retraction groups TPU with PETG as filaments that “may require longer retraction lengths” because they are prone to stringing. Read on its own, that line points you in exactly the direction Prusa and Bambu warn against. The rule is written for filaments that behave like a rigid rod under tension, and TPU does not. Neither Prusa nor Bambu nor Polymaker’s own printer-profile notes for PolyFlex TPU90 publish a retraction number that is supposed to travel between printers, and that is not an oversight. The number that works depends on how far the drive gear sits from the melt zone, how much tube the filament has to buckle into before that, and which hardness you are running, all at once. OrcaSlicer’s own Retraction Test exists for exactly this reason: print it on your machine with your spool rather than copying a figure from a forum thread written for different hardware.

Hardness decides more than any slicer setting

Shore hardness is the number printed on the spool that tells you how soft the material is, and on TPU it changes more about how the print behaves than temperature or speed. Bambu Lab ranks common grades from hardest to softest as 77D, 55D, 95A, 85A, 83A, 80A, 75A, then 70A, and treats them as three different tiers of printer compatibility rather than one continuous scale. The 77D and 55D grades are hard enough to run through Bambu’s own AMS. The 95A down to 80A range prints fine directly but can bend or jam inside the AMS specifically, which is a feeding problem rather than a print-quality one. Below that, at 75A and 70A, Bambu’s own documentation says do not run these through its printers at all: the extruder wheel flattens filament that soft before it can feed, and a flattened section clogs rather than extrudes.

Prusa’s guidance is less categorical. Its flexible materials article gives a wider working range, roughly 60A to 90A, and sums up the relationship in one line: the softer the filament, the harder it is to print. No hard cutoff, just a warning that difficulty climbs as the number drops. Between the two, the practical takeaway is the same. If a spool does not list a Shore hardness at all, that is the first thing to track down, before touching retraction or temperature.

Direct drive versus Bowden changes what "minimal" retraction means

A direct-drive extruder puts the drive gear a few millimeters from the melt zone, leaving almost nowhere for a compressible filament to buckle before it actually feeds. A Bowden setup runs the filament through a meter or more of PTFE tube first, and that tube gives TPU room to bow and coil under the exact compression that causes the jamming described above, on top of whatever happens right at the gear. Polymaker’s own printer-profile page for PolyFlex TPU90 does not hedge on this: it states plainly that the material is not advisable to print on a Bowden setup at all, separate from any retraction or speed tuning.

That does not mean every Bowden printer is off the table for every TPU, since a stiffer 95A grade behaves very differently from a soft 80A one in the same machine. It does mean a retraction setting copied from a build thread is only as useful as the extruder layout it was written for. The same number that barely strings on a direct-drive setup can jam outright on a Bowden printer, and the fix in that case is not more retraction. It is accepting that the tube length changes the ceiling on what retraction can safely do, and dropping expectations, or the print speed, to match.

Temperature and cooling: the one place TPU runs backward from PLA

On PLA and PETG, dropping nozzle temperature by a few degrees usually reduces stringing, because less molten plastic sits at the tip ready to ooze across a gap. Prusa’s guidance for flexible filaments points the other way: increase nozzle temperature by roughly 5°C above what you would normally run, because that lowers the filament’s resistance as it feeds, and keep the part-cooling fan off entirely. Prusa’s general range for flexible materials sits at 230°C to 245°C, though your specific spool’s data sheet should win over that if the two disagree.

The reason the usual logic flips is that TPU’s stringing risk comes more from feed resistance and elastic recoil in the drive path than from excess heat sitting at the nozzle tip. A slightly hotter melt feeds with less resistance and gives the filament less reason to compress and buckle in the first place, which matters more here than the small amount of extra ooze the higher temperature adds. It is a real trade-off, not a free win, which is part of why treating TPU settings as a copy of your PLA profile with the fan turned down does not work.

Give the ooze somewhere to go

Bambu Lab’s own TPU guide names travel distance directly as a driver of stringing on this material, and its fix is not a slicer setting at all: avoid placing several small models on one plate and printing them layer by layer, and print by object instead, or just print one part at a time. Fewer long travel moves between separate, disconnected islands means fewer chances for ooze to stretch into a visible strand across open air. If your slicer’s combing or “avoid crossing perimeters” option is already doing useful work for stringing on other materials, as covered in Reduce Stringing, it still applies here. It just will not compensate for a plate arranged the way you would arrange a PLA batch print.

Moisture still counts, maybe more here

TPU is hygroscopic enough that Prusa lists it as a standing con of the whole flexible-material category, and Bambu Lab publishes specific drying numbers for it: 65°C to 75°C for 8 hours in a forced-air dryer, or 80°C to 90°C for 12 hours using an X1 or X1C heated bed as the source, flipping the spool every 6 hours either way. If you suspect wet filament generally, from popping sounds to a rough, bubbled strand, the full diagnostic sequence is in Wet Filament Symptoms.

Worth saying plainly: Bambu’s own documentation does not promise that drying fixes this. It states that drying reduces stringing, not that it eliminates it, especially on a plate with a lot of travel between features. That is a more honest bar than most stringing advice sets, and it is worth carrying into how you judge a print rather than chasing a setting that is supposed to make the problem disappear entirely.

What to do about the strings that are still there

Once temperature, retraction, plate layout, and moisture are all reasonable, some webbing on TPU is normal rather than a sign something is still wrong. Bambu’s own fix for leftover strings is post-processing, not further slicer tuning: warm air from a hairdryer to shrink them back, or a lighter’s flame passed quickly near the surface to burn them off, with an obvious caution to keep skin and the model itself clear of the flame. Treating a print with faint, easily removed webbing as finished, rather than as a failure that needs another round of retraction changes, is the realistic stopping point for this material.

TPU stringing checklist

• Confirm the spool’s Shore hardness before changing any setting. 95A and harder behaves very differently from 80A and softer.
• On Bambu printers, keep 75A and softer grades out of the AMS and off the printer entirely per Bambu’s own guidance; run 95A to 80A direct rather than through the AMS.
• Start retraction at zero or your printer’s stock TPU profile, then raise it in small steps while watching for grinding, not the other way around.
• Use a direct-drive extruder where you can. If you are on Bowden, expect a stiffer grade to behave better than a soft one, and do not assume a forum’s retraction number will transfer.
• Increase nozzle temperature by roughly 5°C above your normal profile and keep the part-cooling fan off, opposite of what helps on PLA or PETG.
• Print one object at a time, or use “print by object,” instead of batching several small models on one plate.
• Dry the spool (65 to 75°C for 8 hours, or 80 to 90°C for 12 hours on a heated bed) if you notice popping, bubbling, or rough extrusion, and expect this to reduce stringing rather than remove it completely.
• Clean up remaining strings with warm air or a brief pass of flame rather than chasing them with more retraction.