Sententiam Ferre

A mobile-first collaborative decision-making platform with swipe-style voting. The name comes from Latin, meaning "to cast a vote" or "to pass judgment."

Comment

Sententiam Ferre - BRIA FIBO Hackathon Submission

Inspiration

Game studios face a common challenge: creating consistent 2D character assets requires either expensive artists or unpredictable AI prompt engineering. We wanted to solve this by giving teams deterministic control over character generation while enabling collaborative decision-making. The name "Sententiam Ferre" (Latin for "to cast a vote") reflects our core mission—democratizing creative decisions through structured AI generation.

FIBO's shift from text prompts to structured JSON control was the perfect foundation. Instead of hoping a prompt produces the right result, teams can now specify exact parameters and get consistent, reproducible characters every time.

What it does

Sententiam Ferre is a collaborative character creation platform for 2D mobile game development. It combines BRIA's FIBO API with group-based voting to streamline the character design workflow:

  • Step-by-Step Character Wizard: A guided 7-step wizard walks users through character creation, showing the FIBO JSON being built in real-time. Users select art style, camera angle, pose, expression, color palette, and background—each choice immediately reflected in the structured parameters.

  • Real-Time FIBO Preview: As users make selections, they see exactly what JSON will be sent to FIBO. This transparency demystifies AI generation and teaches users how structured control works.

  • Group Collaboration: Teams create groups, invite members, and collectively decide on character designs through voting. Admins can lock specific parameters to maintain visual consistency across a project. USERS CAN CREATE VARIATIONS AND THE NEW ITEMS GO STRAIGHT BACK INTO THE DECISION DECK FOR EVEYRONE TO VOTE ON.

  • Character Gallery: Generated characters appear in a gallery with status tracking, variation creation, and export capabilities.

    • EXPORT: Download the final image, export the final JSON that everyone (or the quorum you have configured) has agreed upon! NO NEED TO LEAVE COMMENTS ON A FIGMA/MIRO board to constantly iterate over small details and have multiple team mates request multiple small changes to just "see how it looks".

How we built it

Frontend: React with a custom state machine for the wizard flow. The wizard maintains choices and builds FIBO-compatible JSON in real-time, displaying it alongside the selection UI.

Backend: Django REST Framework handling authentication, group management, and generation job processing. The BRIA client uses sync mode for immediate results.

FIBO Integration: We map user-friendly choices (like "Cartoon" or "Pixel Art") to FIBO's structured parameters. The prompt builder combines the character description with style modifiers optimized for 2D game assets.

Architecture:

  • CharacterWizard.jsx - 7-step wizard UI with live JSON preview
  • characterWizardMachine.js - Step configuration and FIBO parameter mappings
  • prompt_builder.py - Converts wizard choices to optimized FIBO prompts
  • generation.py - Job processor handling BRIA API communication

Challenges we ran into

  • Content Moderation: BRIA's content filter is strict. Terms like "attacking" or other terms triggered rejections. We rewrote our prompt modifiers to use family-friendly alternatives while preserving the intended meaning.

  • Parameter Mapping: Translating intuitive UI choices to FIBO's structured format required careful mapping. We needed to ensure the frontend wizard, backend serializer, and prompt builder all stayed in sync.

  • Real-Time Preview: Showing the FIBO JSON updating live as users click through the wizard required careful state management to keep the preview accurate without performance issues.

Accomplishments that we're proud of

  • Educational UX: The real-time JSON preview teaches users how FIBO works. They see the direct connection between their choices and the structured parameters—demystifying AI generation.

  • Production-Ready Workflow: This isn't a demo. It's a complete collaborative tool with authentication, group management, voting, and export capabilities that a real game studio could use. It can become a SaaS application for consensus on Construction plans, other graphic assets, etc.

  • Deterministic Control Showcase: By showing the exact JSON being sent, we demonstrate FIBO's core value proposition: same parameters = same style, every time.

What we learned

  • FIBO's structured approach fundamentally changes how teams can work with AI generation. Instead of prompt engineering being a dark art, it becomes a predictable, shareable configuration.

  • The wizard pattern is powerful for complex parameter selection. Breaking down FIBO's many options into digestible steps makes the technology accessible to non-technical team members.

  • Content moderation requires careful prompt engineering even with structured parameters—the text description still matters.

What's next for Sententiam Ferre

  • Expanded FIBO Parameters: Add FOV sliders, lighting intensity controls, and composition grids to showcase more of FIBO's capabilities.

  • Template Library: Pre-built character templates (hero, villain, NPC archetypes) that teams can use as starting points.

  • Export Integration: Direct export to game engines with proper sprite sheet formatting.

  • ADDITIONAL USE CASES: Extend beyond game characters to product mockups, marketing assets, and other collaborative visual decisions.

Built With

Share this project:

Updates