Guides for building, testing, and shipping

Choose an approach, prototype a feature, and keep long-running AI work away from the frame loop.

• Game AI without machine learningFinite state machines, behavior trees, and utility AI cover most shipped game AI. Here is how each one works, when to use it, and where it breaks down.
• Adding LLM-driven dialogue to NPCsThe production tradeoffs for LLM-driven NPC dialogue: local vs cloud costs, keeping NPCs in character, and the four failure modes that block a ship.
• Zero to a talking NPC in 30 minutesInstall Ollama, pull a small local model, and wire one self-contained HTML file to its API to get a character talking back to you, no account, no cloud.
• Shipping AI features without wrecking performanceThe budget mindset for shipping AI in games: frame time vs request latency, quantization tradeoffs, on-device memory limits, and cloud cost math.

Use a repeatable visual pipeline, understand model and asset licences, and keep a record before release.

• Keeping AI-generated 2D art style-consistentStyle blocks, seeds, reference-image conditioning, and LoRA training for 2D game art, explained plainly with an honest read on when each earns its setup cost.
• The text-to-3D pipeline that actually reaches an engineText-to-3D generation produces a blockout, not a shippable mesh. The real pipeline: inspect, retopologize, re-UV, texture, rig, import at a consistent scale.
• AI audio for games: music, SFX, and voiceAI audio for games: full tracks versus loopable stems, why layered short SFX generations beat one-shot prompts, and the hard consent line on voice cloning.
• AI game assets: copyright, player trust, and pipeline disciplineA field guide to AI game assets: copyright by jurisdiction, why players bounce off careless AI art, a pipeline for visual coherence, and rights-clear sources.
• Open-weight model licenses for game developers, comparedA checked comparison of Llama, Mistral, Qwen, Gemma, Stable Diffusion, and FLUX licenses for indie game developers: what ships commercially and what does not.
• Reading AI tool terms before you shipHow to read AI tool terms as a game developer: what commercial use means, who owns generated output, training-data clauses, red flags, and a pre-ship checklist.

Prepare a build for feedback, make a focused request, and use a jam or public post to find the next issue.

• How to give feedback that helpsGive useful game feedback with the version, device and exact moment. Keep discussion in the existing Discord thread and protect private player information.
• The jam handbook: entering and running a jamEnter a jam with your existing game thread and an external build. Follow the event rules, moderator site approval, Discord reaction voting and results workflow.
• Posting your game to ShowcaseFind the right place for playable prototypes, real game development and visual experiments; avoid repetitive promotion and request focused feedback.

Review platform rules, player-data implications, and tool terms before a build reaches players.

• Platform rules on AI content: Steam, itch, Epic, consolesWhat Steam, itch.io, Epic, Xbox, PlayStation, Nintendo, and mobile stores require you to declare about AI-generated content in your game, checked August 2026.
• Player data and cloud AI: what leaves the deviceWhat travels off the device when an NPC calls a cloud AI API, what you must disclose, how to cut what you send, and where age and region rules apply.
• Reading AI tool terms before you shipHow to read AI tool terms as a game developer: what commercial use means, who owns generated output, training-data clauses, red flags, and a pre-ship checklist.
• AI game assets: copyright, player trust, and pipeline disciplineA field guide to AI game assets: copyright by jurisdiction, why players bounce off careless AI art, a pipeline for visual coherence, and rights-clear sources.