Boogu-Image-0.1: open-source model series with top bilingual text rendering
10B parameter model does dense text and photorealism with 3-4 step Turbo mode.
Boogu-Image-0.1 is a new open-source model series from Boogu, offering unified image generation and editing under the Apache-2.0 license. The family includes three variants: Boogu-Image-0.1-Base for dense text rendering (ideal for posters, documents, and complex bilingual designs at 2K resolution), Boogu-Image-0.1-Turbo for high-quality photorealism in just 3–4 inference steps while preserving bilingual text rendering, and Boogu-Image-0.1-Edit for image editing and transformation. All models have 10B parameters and require 12–80GB VRAM depending on configuration.
Despite using roughly one order of magnitude less training data than existing open-source models, Boogu-Image-0.1 delivers competitive performance in text-to-image generation, fast sampling, and editing. The project emphasizes systematic improvements in understanding ability, data quality, and the training pipeline. For workloads dominated by dense text, the Base model at 2K output is recommended; for photorealism, Turbo is the better default. The models are available on Hugging Face, with a GitHub repository and ComfyUI support for easy integration.
- Three variants: Base (text rendering), Turbo (3-4 step fast photorealism), Edit (image editing).
- 10B parameter model trained on ~10× less data than comparable open-source models.
- Strong bilingual Chinese-English text rendering suitable for posters, documents, and brand guides.
Why It Matters
Open-source image generation with reliable multilingual text rendering lowers barriers for design and publishing.