Introducing Bria FIBO on WaveSpeedAI

Introducing Bria FIBO: The First JSON-Native Text-to-Image Model on WaveSpeedAI

WaveSpeedAI is excited to announce the availability of Bria FIBO, a groundbreaking text-to-image model that brings unprecedented precision and control to AI image generation. FIBO represents a fundamental shift in how developers and enterprises approach visual content creation—moving from unpredictable prompt engineering to structured, reproducible, and legally compliant image generation.

What is Bria FIBO?

FIBO is an 8-billion parameter state-of-the-art text-to-image model developed by Bria AI, and it’s the first model of its kind to be truly JSON-native. Unlike traditional text-to-image models where outputs can vary unpredictably, FIBO converts your prompts into structured JSON schemas that precisely specify every visual element—from lighting and composition to camera settings and depth of field.

Built on a DiT-based flow-matching architecture with SmolLM3-3B as the text encoder and Wan 2.2 as the VAE, FIBO was trained exclusively on licensed data with structured JSON captions averaging 1,000+ words each. This unique training approach enables the model to understand and respond to nuanced visual instructions with remarkable accuracy.

What truly sets FIBO apart is its enterprise-grade foundation: 100% licensed training data, full GDPR and EU AI Act compliance, complete provenance tracking, and C2PA watermarking. For businesses navigating the complex landscape of AI-generated content rights, FIBO eliminates legal barriers entirely.

Key Features

• JSON-Native Architecture: Every prompt is converted into a machine-readable structured schema, enabling versioning, automation, and seamless integration into complex production workflows

• Three Powerful Operating Modes:

  • Generate: Transform short prompts into comprehensive JSON schemas with 1,000+ word descriptions covering lighting, camera, composition, and more
  • Refine: Make targeted attribute changes (e.g., “85mm lens,” “warmer tones”) without affecting unrelated elements—no more prompt drift
  • Inspire: Extract structured prompts from reference images and blend them with new creative intent for consistent variations

• Disentangled Control: Tweak individual attributes like camera angle, lighting direction, or color palette without breaking the overall scene composition

• Enterprise-Ready Compliance: Built on 100% licensed data with full legal indemnification, GDPR compliance, and EU AI Act readiness

• Strong Prompt Adherence: FIBO outperforms comparable open-source baselines on PRISM-Bench evaluations for both text alignment and aesthetic quality

• Reproducible Outputs: Same prompt, parameters, and seed always produce identical results—essential for brand consistency and production pipelines

• VLM-Guided Prompting: Integrates vision-language models to transform brief descriptions into rich, structured schemas automatically

Use Cases

Brand and Marketing Assets

Create consistent visual campaigns across channels. FIBO’s structured approach ensures your product shots maintain the same lighting style, background treatment, and brand aesthetic whether you’re generating one image or one thousand. The reproducibility feature is invaluable for A/B testing where you need to change only specific variables.

E-Commerce Product Photography

Generate professional product imagery at scale. Define your studio setup once in a structured prompt—lighting configuration, camera angle, background—then swap products while maintaining perfect consistency. This dramatically reduces the cost and time of traditional product photography.

Social Media Content Creation

Produce platform-optimized visuals with built-in aspect ratio support for Instagram squares (1:1), Stories and Reels (9:16), or YouTube thumbnails (16:9). The negative prompt feature helps you avoid common artifacts like text overlays, watermarks, or distorted elements.

Advertising and Campaign Development

The Refine mode shines for iterative creative development. Start with a concept, then progressively adjust elements—“make it sunset lighting,” “shift to a 35mm wide angle,” “add depth of field”—while preserving your core composition. Each iteration is traceable and reproducible.

Automated Content Pipelines

FIBO’s JSON-native design makes it ideal for programmatic content generation. Feed structured specifications from your CMS, DAM, or e-commerce platform directly into the API. The machine-readable schemas integrate naturally with existing automation infrastructure.

Concept Art and Visual Development

Explore creative directions rapidly using the Inspire mode. Upload a reference image, and FIBO’s vision-language component extracts a detailed structured description that you can then modify and regenerate—perfect for exploring variations on a theme.

Getting Started on WaveSpeedAI

Accessing Bria FIBO through WaveSpeedAI is straightforward and offers several advantages over self-hosting or other platforms:

1. Visit the Model Page: Navigate to https://wavespeed.ai/models/bria/fibo to access the model directly

2. Craft Your Prompt: Start with a clear description of your subject, setting, lighting, and style. For example: “A cozy rooftop garden in Paris at night, warm string lights, shallow depth of field, cinematic glow”

3. Select Your Aspect Ratio: Choose from 1:1 (social square), 16:9 (landscape/banners), 9:16 (portrait/reels), 4:5, or 3:2 depending on your output needs

4. Refine with Optional Controls:

   • Add a negative prompt to avoid unwanted elements (“blurry, extra fingers, watermark, text artifacts”)
   • Use structured_prompt for schema-style guidance with specific object lists or layout constraints
   • Set a seed for reproducible results

5. Generate: Run the model and receive your image in seconds

At just $0.04 per image, FIBO on WaveSpeedAI delivers exceptional value. Combined with our platform’s zero cold starts and consistently fast inference times, you can integrate production-quality image generation into your workflows without infrastructure headaches or unpredictable latency.

Why WaveSpeedAI?

Running FIBO through WaveSpeedAI means you get:

• No Cold Starts: Your API calls are processed immediately without waiting for model loading
• Consistent Performance: Reliable inference times for production applications
• Simple Integration: RESTful API that integrates with any tech stack
• Affordable Pricing: Pay only for what you use at competitive per-image rates
• Scalability: Handle burst traffic without provisioning infrastructure

Conclusion

Bria FIBO represents a new paradigm in text-to-image generation—one where precision, reproducibility, and legal clarity are built into the foundation rather than bolted on as afterthoughts. Whether you’re building automated content pipelines, creating marketing assets at scale, or developing products that need consistent visual generation, FIBO’s JSON-native architecture provides the control and predictability that professional workflows demand.

The combination of state-of-the-art image quality, enterprise-grade compliance, and structured control makes FIBO an essential tool for teams serious about integrating AI image generation into their products and processes.

Ready to experience structured image generation? Try Bria FIBO on WaveSpeedAI today and discover what’s possible when AI image generation becomes truly controllable.