FIBO

Bria-AI活跃维护

首个开源的 JSON 原生文生图模型，用结构化 JSON 输入实现可控、可预测且版权合规的图像生成。

FIBO is a SOTA, first open-source, JSON-native text-to-image model built for controllable, predictable, and legally safe image generation.

暂无版本359star22fork同步于 2026-11-10 12:01 UTC

文生图 JSON 原生可控生成合规数据图像生成

对比 GitHub 官网 agentic-ai ai controlable-image-generation creative-ai deep-learning

怎么上手

由 README 提炼，执行前请确认

原文

git clone https://github.com/Bria-AI/FIBO.git && cd FIBO
uv sync
source .venv/bin/activate && export PYTHONPATH=${PYTHONPATH}:${PWD}

硬件要求按 README 说明

README 未说明显存与 GPU 型号要求；模型为 8B 参数，本地推理需自行准备环境，也可改用云端 API 或 ComfyUI 节点。

支持 Apple Silicon

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼

没人工点评？写下你的经验

首个开源 JSON 原生文生图模型，用结构化长描述实现可控可复现的图像生成

首个开源的 JSON 原生文生图模型，用结构化 JSON 输入实现可控、可预测且版权合规的图像生成。

核心能力

用 VLM 把短提示扩展成 1000+ 词结构化 JSON 提示
支持 Refine 模式只改指定属性而不破坏整体场景
支持以图片为灵感反推结构化提示再生成变体
启用 TeaCache 可加速约 3 倍且质量损失极小
提供 ComfyUI 节点与 Bria、Fal.ai、Replicate API 接入

容易踩的坑

需先在 Hugging Face 申请模型访问权限并执行 hf auth login；默认用 Gemini 作 VLM 需设置 GOOGLE_API_KEY，否则要加 --model-mode local 用本地 VLM；开源权重仅限非商业使用，商用需另行联系 Bria。

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

用户点评

登录用户的真实使用记录，可按版本和硬件筛选

共 0 条

还没有人点评。第一个说出真实体验的人最有价值。

README 原文

未翻译

在 GitHub 打开

Development

This project uses a Makefile to streamline common development tasks.

To install dependencies and set up pre-commit hooks, run:

make install

The following commands are also available:

make lint: Run linters to check for code quality.
make format: Format the code according to the project's style guidelines.
make check: Run both linters and formatters.
make clean: Remove the virtual environment.
make help: Display a list of all available commands.

python generate.py --prompt "A hyper-detailed, ultra-fluffy owl sitting in the trees at night, looking directly at the camera with wide, adorable, expressive eyes. Its feathers are soft and voluminous, catching the cool moonlight with subtle silver highlights. The owl's gaze is curious and full of charm, giving it a whimsical, storybook-like personality." --seed 1 --output examples/outputs/generate.png

python generate.py --prompt "A hyper-detailed, ultra-fluffy owl sitting in the trees at night, looking directly at the camera with wide, adorable, expressive eyes. Its feathers are soft and voluminous, catching the cool moonlight with subtle silver highlights. The owl's gaze is curious and full of charm, giving it a whimsical, storybook-like personality." --seed 1 --output examples/outputs/generate.png --model-mode local

python generate.py --structured-prompt examples/outputs/generate.json --prompt "make the owl brown" --output examples/outputs/refine.png

python generate.py --image-path assets/original.png --output examples/outputs/inspire.png

python generate.py --image-path assets/original.png --prompt "Make futuristic" --output examples/outputs/inspire-prompt.png

python generate.py --prompt "your prompt" --enable-teacache

python generate.py --prompt "your prompt" --enable-teacache --teacache-threshold 0.8

Citation

We kindly encourage citation of our work if you find it useful.

@misc{gutflaish2025generating,
  title         = {Generating an Image From 1,000 Words: Enhancing Text-to-Image With Structured Captions},
  author        = {Gutflaish, Eyal and Kachlon, Eliran and Zisman, Hezi and Hacham, Tal and Sarid, Nimrod and Visheratin, Alexander and Huberman, Saar and Davidi, Gal and Bukchin, Guy and Goldberg, Kfir and Mokady, Ron},
  year          = {2025},
  eprint        = {2511.06876},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV},
  doi           = {10.48550/arXiv.2511.06876},
  url           = {https://arxiv.org/abs/2511.06876}
}