SD-Trainer-Plugin

Openplanet plugin for Trackmania that swaps skidmark textures in real time based on speed drift quality.

Download

• Get the latest build from the Releases page: https://github.com/TNTisdial/sd-trainer-plugin/releases/latest

Install

• Copy SD Trainer.op to C:\Users\<username>\OpenplanetNext\Plugins.
• Launch Trackmania with Openplanet, change Signature Mode to Developer
• In Openplanet click the developer tab, hover over Load Plugin
• Click SD-Trainer-Plugin

What it does

• Computes drift quality (driftQualityRatio) each frame from measured acceleration vs surface-specific expected max acceleration.
• Maps quality to tiers: default, poor, mid, high.
• Swaps skid textures across asphalt, dirt, and grass.
• Includes a skin-picker UI for selecting tier textures per surface.

Documentation

• docs/io-system.md for staging and swap pipeline details.
• docs/acceleration-logic.md for drift-quality computation.
• docs/settings-reference.md for runtime settings and defaults.
• docs/skid-picker-ui.md for UI behavior and hidden persisted fields.
• docs/code-architecture.md for module ownership and execution flow.

Build package

• Run python3 build_op.py from repo root.
• The script creates an auto-slotted build artifact and also updates SD Trainer.op to the newest build.
• For releases and local install, use SD Trainer.op.

Credit

Thanks to the people and projects that made this possible:

• XertroV (XertroV/tm-modless-skids) for core modless skid swap reference patterns.
• voblivion (voblivion/Openplanet-IMG) for DDS preview helper code used in DDS_IMG/.
• Magpie (SilasDo/Trackmania-Speed-Drift-Trainer-Plugin) for speed-drift acceleration logic direction.
• Runes for testing
• Shorty for proving the concept.

This project was passionately written using Zed editor and OpenCode.

AI Models such as GPT-5.3 Codex and Opus 4.6 had a hand in project-wide changes, like refactoring, architectural changes, documentation updates, packaging and implementation ideas.

Every line generated by AI, via tab completion and agentic loops have been thoroughly evaluated, to comply with Openplanet's AI generation policy.

This project was not vibe coded, but AI helped.

Taylor