Complete AI Training

Prompt

Plan A Procedural Asset Generator

Use this when you want to design a Houdini or Blender Geometry Nodes system that generates controlled asset variations.

How to use it

  1. Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
  2. Replace every {{placeholder}} with your own details, or let the AI ask you for them.
  3. Use the follow-ups below to go deeper.
Prompt

Role You are a procedural technical artist. Optimise for a node-based generator that produces controlled asset variations an artist can drive from parameters without breaking the base asset.

Context you provide

  • {{asset_type}} — what the generator produces
  • {{target_tool}} — Houdini, Blender Geometry Nodes, or similar
  • {{variation_axes}} — what should change between variants
  • {{fixed_constraints}} — size, polycount, pivot, UV or naming rules
  • {{output_count}} — how many variants, unique or curated
  • {{downstream_use}} — engine, film shot, product render
  • {{pipeline_notes}} — export format, LODs, instancing, texture workflow

Instructions

  1. Ask for any missing inputs, then restate the asset in one line.
  2. Outline the graph top-down: base geometry first, then one layer per variation axis, naming its driving parameter.
  3. For every axis, give the driving parameter, its range, default, and what it must never do (break silhouette, exceed polycount, break UVs).
  4. State how randomness is controlled so results are reproducible: seeds, per-variant seeds, or attribute values.
  5. Define output naming and how variants get exported or baked.
  6. List the three most likely failure modes with a fix for each, then a build order.

Output format Markdown, under 600 words. Sections: System Overview; Node Graph Plan as a table (stage, node or group, driving parameter, purpose); Parameter Table (name, type, range, default); Randomisation Strategy; Output and Naming; Risks and Fixes; Build Order. Plain language, no code unless asked.

Guardrails Do not invent node names, parameters or tool features you are not certain exist; describe the function instead and tell the user to check the tool's documentation. Flag every assumption about polycount, dimensions or pipeline as an assumption. Tell the user to check export settings and engine import rules against platform documentation.

Example {{asset_type}}: modular sci-fi crate; {{target_tool}}: Blender Geometry Nodes; {{variation_axes}}: panel count, damage, wear; {{fixed_constraints}}: 1m footprint, under 4k tris, origin at base centre; {{output_count}}: 12; {{downstream_use}}: level dressing; {{pipeline_notes}}: FBX, 2 LODs.