KlipperLearn

FREE · OPEN SOURCE · COMMUNITY PREVIEW

A smarter workflow for the printer you already own.

Turn original photographs, print logs and repeatable trials into reviewed slicer settings. Keep your machine. Understand every adjustment.

Explore the previewChoose your setupTry the reference demo

Demo: original 0.1.0 local evaluator, not the current host application. Phone optional. GPL-3.0-or-later. No publisher subscription. No guaranteed performance gains.

Concept illustration of a legacy printer with a phone companion
AI-generated concept illustration, not a hardware-validation photograph.
Experimental, supervised software. The file-first optimizer creates proposals and files; it does not start a printer. Existing KlipperLearn host controls are a separate, authenticated setup. Review settings, keep originals, and supervise physical tests.

ONE CONFIGURATION, THREE PRIORITIES

Compare evidence. Not promises.

01

Quality

Prioritize reviewed surface and geometry scores from repeatable, acceptable trials.

02

Standard

Balance quality and time with an explicit, inspectable selection policy.

03

Speed

Choose the fastest acceptable tested configuration without silently reducing layer count.

Modes can legitimately be identical. Missing or incomparable evidence produces a clear explanation—not invented settings or a claimed optimum.

START SMALL

Pick only the tools you need.

All packages are free. The release includes source, licenses and SHA-256 checksums. These are independently maintained integrations, not upstream endorsements.

Community preview release ↗Versioned downloads, source code, checksums and validation notes. File-first AI optimizerPortable skill, local Python helper and original calibration STL. Public ChatGPT directory publication is still pending. Ecosystem packagesCura, OctoPrint, Orca/PrusaSlicer and a standalone evidence toolkit. Phone and local-host companionUse your own Klipper/Moonraker host for the web console, camera and available phone sensors.

Three steps, with a human in control.

  1. Identify the setup. Keep printer, nozzle, material, slicer and benchmark identity explicit.
  2. Print and review comparable trials. Preserve original images, failures, layer counts, timing and independent repetitions.
  3. Export, inspect and re-slice. Apply a reviewed preset only after checking the effective settings and slicer preview.

For OrcaSlicer, modes use paired process and filament presets. Other slicers currently use evidence adapters or a settings report, not a universal preset format.

WORK WITH THE ECOSYSTEM

Ten integration paths. One transparent status.

Klipper Read-only diagnosticsMoonraker History and evidenceOrcaSlicer Paired preset auditsPrusaSlicer G-code fingerprintsCura Evidence scriptOctoPrint Independent evidence pluginMainsail Camera interoperabilityFluidd Camera interoperabilityKIAUH Host coexistence checksVoron Physical-test evidence review

External tooling—not ten merged features. Native application and physical acceptance requirements are recorded in the integration ledger.

Reproducibility comes first.

Keep different printers and test models separate. The Flashforge Benchy series was operator tuning advised by ChatGPT, not KlipperLearn. The Anycubic 4Max Pro charts and cube trials used the separate KlipperLearn, phone and webcam setup.

Read the benchmark provenance →

Help close the real gaps.

We welcome native-slicer import checks, reproducible bug reports and safely supervised validation. Share a minimal sanitized example, not credentials or your whole printer workspace.

Report compatibility →
No accounts or tracking on this site. This static page has no scripts, analytics or forms. Tools you choose to use retain their providers' policies. ChatGPT plans, regions and usage limits are controlled by OpenAI; the project cannot promise universal or unlimited access.