JSON-native structured captions
According to the model card, FIBO is trained exclusively on long structured JSON captions up to 1,000+ words for disentangled control.
Open Source Model Profile · briaai
FIBO
FIBO is a text-to-image model from briaai. According to the model card, it is JSON-native and trained on long structured captions, with captured Safetensors metadata reporting about 8.29B parameters.
- Publisher
- briaai
- Task
- text-to-image
- Model type
- Unknown
- License
- bria-fibo
- Library
- diffusers
- Publication status
- Accepted · not indexed
Model overview
FIBO is published by briaai as a text-to-image model. Captured Safetensors metadata reports 8,285,836,848 parameters, or about 8.29B. According to the model card, it is a JSON-native model trained exclusively on long structured captions for controllable, reproducible image generation.
Recorded capabilities
Generate, Refine, Inspire modes
The model card documents three modes: Generate expands short prompts, Refine edits attributes, and Inspire starts from an image to extract and blend structured prompts.
Production and local paths
According to the model card, production paths include API endpoints, ComfyUI Generate and Refine nodes, and local inference.
DiT flow-matching architecture
According to the model card, FIBO is an 8B-parameter DiT-based flow-matching model using SmolLM3-3B text encoding with DimFusion conditioning and a Wan 2.2 VAE.
Use cases in the source record
- Professional text-to-image workflows that start from short ideas and expand them into structured JSON prompts for lighting, composition, color, and camera control, as described in the model card.
- Iterative refinement where a short instruction updates only requested attributes of an existing structured prompt, according to the model card.
- Image-inspired generation where a vision-language step extracts a structured prompt from an input image and blends it with creative intent, according to the model card.
Limitations and unknowns
- The publisher's comparison claim that FIBO outperforms comparable open-source baselines is a model-card claim and has not been independently verified by Ethen.
- The publisher's prompt-adherence claim of high alignment on PRISM-style evaluations is a model-card claim without extracted scores in this record.
- Captured metadata records an other license value; the model card says weights are open source for non-commercial use and directs commercial users to a separate license path.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: briaai/FIBO
Captured: Unknown. Processed: 2026-09-07T19:35:54.502139+00:00.
FIBO is the first open-source, JSON-native text-to-image model trained exclusively on long structred captions. Fibo sets a new standard for controllability, predictability, and disentanglement by implementing the new VGL - Visual GenAI Language paradigm 🌍 What's FIBO? Most text-to-image models excel at imagination—but not control. FIBO is built for professional workflows, not casual use. Trained on structured JSON captions up to 1,000+ words , FIBO enables precise, reproducible control over lighting, composition, color, and camera settings. The structured captions foster native disentanglement, allowing targeted, iterative refineme…
F001F002F003F004F005F006F007F008F009F012F013F014F015F016F017F018F019F028F029F030