Inspiration
Game development is 10 percent coding and 90 percent asset creation. Indie developers struggle the most, needing consistent 2D animations and usable 3D models without the time, cost, or skills of full art pipelines. We saw that most AI tools generate beautiful images, but not assets that actually work inside game engines. GenForge was inspired by the need to bridge this gap and turn AI output into real, playable game assets.
What it does
GenForge is an AI-powered game asset generator that transforms simple text prompts into production-ready 2D sprite sheets and 3D models. Using Bria FIBO’s JSON-native structured prompts, it generates consistent multi-frame animations, engine-ready sprite sheets with metadata, and textured 3D models exportable as GLB or OBJ files. Everything is built to work directly with Unity, Godot, and Phaser without manual cleanup.
How we built it
We built GenForge as two specialized engines under one unified experience.
The 2D Sprite Engine uses Flask and Bria FIBO to generate full animation sprite sheets in a single call, ensuring character consistency. A post-processing pipeline handles background removal, frame slicing, padding, and metadata generation.
The 3D Engine uses FastAPI with Bria FIBO to create controlled reference images, which are then converted into 3D meshes using Tripo AI. An asynchronous job system manages long-running tasks and keeps the user informed in real time.
A React and Three.js frontend brings both engines together with live previews and downloads.
Challenges we ran into
Maintaining character consistency across animation frames was a major challenge, as AI often changed designs between frames. Background removal was unreliable due to fake transparency and checkerboard patterns. In 3D generation, long processing times caused UX issues, requiring a robust job tracking and polling system. We also had to carefully handle API authentication, rate limits, and credit management without exposing keys to the client.
Accomplishments that we're proud of
We achieved consistent multi-frame sprite animations in a single AI call. Our outputs work directly in popular game engines with zero manual slicing. We successfully unified two very different AI pipelines, 2D sprites and 3D meshes, into one clean developer experience. Most importantly, we turned AI-generated content into truly game-ready assets.
What we learned
We learned that structure matters more than creativity when building professional AI tools. JSON-native prompts consistently outperformed free-form text. The hardest part was not generation, but making assets usable in real workflows. We also learned how critical async systems and clear UX feedback are when dealing with long-running AI tasks.
What's next for GenForge
Next, we plan to add AI-generated texture projection for 3D models, auto-rigging for animation, and higher frame-count sprite animations. We aim to build direct plugins for Unity and Godot, and evolve GenForge into a full text-to-scene pipeline for game developers.
Log in or sign up for Devpost to join the conversation.