Research

Converting filament profiles between slicer dialects

Bambu Lab, Prusa, Polymaker and OrcaSlicer each publish filament settings in their own format — their own dialect. I'm studying whether these can be converted systematically into settings for a different machine. As far as I could find, there is no prior work on this problem.

Method

I analysed more than 7,000 published profiles. The conversion runs in two stages: the source dialect is first rewritten as a physical intermediate representation (volumetric flow, thermal history, bead geometry, cooling rate), and the target dialect is written out from that. Parameters are treated in three classes: material-specific, machine-specific, and the interaction between the two.

How the conversion flows

Input

Source dialect

Published profiles from Bambu Lab / Prusa / Polymaker / OrcaSlicer

Intermediate

Physical quantities

Volumetric flow
Thermal history
Bead geometry
Cooling rate

Output

Target dialect

Settings for the target machine

Material-specific: Set by the filament
Machine-specific: Set by the printer
Interaction: Set by the two together (e.g. cooling fan)

Findings

65% of the variance between machines is explained by a single axis: cooling strategy.
For the same filament, official settings differ by up to 2.5× between ecosystems.
Moving from glass-fibre to carbon-fibre filled material reduces maximum volumetric flow by a factor of 0.76–0.85, regardless of vendor.

How public profiles are distributed

OrcaSlicer · 7,331 profiles

PrusaSlicer · 7,155 profiles

14,486 public profiles counted by nozzle temperature and max volumetric flow; darker means more profiles. 296 fall outside the plotted range. Flow settings spread widely even at the same temperature, and the shape differs by source.

Hardware test (September 2026, PLA)

Print time. The converted profile finishes in about the same time as the unconverted one; only the existing profile takes longer.

Print time in the hardware test (September 2026, PLA)
Profile	Print time
Existing profile	84.9 min
Converted profile	57.7 min
Unconverted (source dialect as is)	57.0 min

More than 80% of the time difference came from different cooling thresholds. Quality was about the same, except on overhangs, where the existing profile did better. The cause was values that assume an auxiliary fan being copied across as they were. I changed the approach to treat fan values as an interaction term that depends on the machine.

Presentation and next steps

I gave an oral presentation at Ginnan Gakkai 2026. Testing continues with ASA, PETG and PPS-GF20.