Good day Product Hunt,
LLMs are strangely dangerous at making actual video video games. Ask one for a 3-D platformer and you can get a unmarried HTML document with 800 traces of hand-rolled math seeking to pretend physics, a render loop held along side duct tape, and collisions that do not paintings. The reason being easy: the AI is not sitting on best of a recreation engine. Each body of capability — rendering, physics, ECS, animation, shadows, collision, networking — needs to be re-invented from scratch within the suggested. That does not scale previous a toy video games.
ParallaxPro fixes that through tightly integrating an AI assistant with a browser-based 3-D recreation engine.
As an alternative of asking the LLM to be the engine, we give it one. The AI simply puts entities, attaches habits scripts, and wires up programs — the engine handles the remainder:
– WebGPU rendering — trendy graphics, runs within the browser
– Inflexible frame physics — powered through Rapier
– Skeletal animation — actual rigged characters
– Entity-Part-Device structure — data-oriented, no longer OOP
– 5000+ 3-D belongings — Kenney, Poly Haven, and extra, already listed for the AI to make use of
– One-click put up to https://parallaxpro.ai/video games/
And the entire stack is absolutely open supply — engine, editor, device activates, LLM compiler, recreation templates. No hidden black packing containers, no engine royalties, no dealer lock-in. Convey your individual LLM (Claude, GPT, Gemini, Groq, Ollama, Codex, Copilot) or simply use the hosted model and get started typing.
Kind “chess” and get chess. Kind “fps shooter” and get an fps shooter. Then stay chatting to mend insects, upload options, song for your liking — the AI reads your recreation code, edits scripts, and validates adjustments.
Check out it utterly unfastened: https://parallaxpro.ai
Repo: https://github.com/ParallaxPro-A…
Would like your comments, worm studies, and wildest suggested concepts.
