Abiray/HiDream-O1-Image-FP8

HiDream-O1-Image (FP8 Scaled)

Repository Notice: This repository hosts the FP8 scaled (hidream_o1_image_fp8_scaled.safetensors) quantized weights of the base 8B HiDream-O1-Image model. This standard 8-bit quantization halves the file size and memory footprint compared to the uncompressed bf16 model, making it highly optimized for fast inference on mid-range GPUs (12GB - 16GB VRAM) while preserving exceptional visual fidelity.

For the official, uncompressed weights and source code, visit the original HiDream-O1-Image repository.

---

HiDream-O1-Image is a natively unified image generative foundation model built on a Pixel-level Unified Transformer (UiT) without external VAEs or disjoint text encoders, which natively encodes raw pixels, text, and task-specific conditions in a single shared token space — supporting text-to-image, image editing, and subject-driven personalization at up to 2,048 × 2,048.

Key Features

• 🧬 Pixel-Level Unified Transformer — One end-to-end model on raw pixels, no VAE, no disjoint text encoder.
• 🎨 One Model, Many Tasks — Text-to-image, long-text rendering, instruction editing, subject-driven personalization, and storyboard generation in a single architecture.
• 🧠 Reasoning-Driven Prompt Agent — Built-in "thinking" agent that resolves implicit knowledge, layout, and text rendering before generation.
• 🖼️ Native High Resolution — Direct synthesis up to 2,048 × 2,048 with sharp fine-grained detail.
• ⚡ Exceptional Efficiency and Versatility at 8B Scale — With only 8B parameters, achieves performance parity with or even surpasses larger open-source DiTs and leading closed-source models.

Usage & Installation

To use these FP8 weights, you will need the original HiDream inference pipeline.

1. Install the Original Repository

Clone the official repository and install dependencies:

git clone [https://github.com/HiDream-ai/HiDream-O1-Image.git](https://github.com/HiDream-ai/HiDream-O1-Image.git)
cd HiDream-O1-Image
pip install -r requirements.txt