HiDream-O1-Image-1.5 - A commercial image generation model launched by HiDream.ai

What is HiDream-O1-Image-1.5

HiDream-O1-Image-1.5 is a commercialimage generationlarge model launched by HiDream.ai, based on the native omni-modal UiT architecture. On the Artificial Analysis text-to-image leaderboard, it ranks third globally and first in China with an ELO of 1265, surpassing Google Nano Banana 2 and ByteDance Seedream 4.0. The model features photographic-grade portraits, fine-grained animal modeling, precise text rendering, and multi-subject consistency. It targets commercial scenarios such as advertising, brand design, e-commerce visuals, and film storyboarding, marking HiDream.ai's firm position in the global top tier of visual generation.

Main features of HiDream-O1-Image-1.5

• Portrait photography generation: Supports magical lighting, two-person interaction, and character close-ups, with natural performance in skin texture, clothing texture, body relationships, and environmental blur.
• Animal and natural environment generation: Fine-grained modeling of animal structures, fur texture, dynamic performance, as well as complex lighting, underwater refraction, and other scenes.
• Text rendering and layout: Possesses accurate text generation capabilities and complex layout capabilities.
• Multi-subject consistency: Supports coordinated generation and visual storytelling of multiple characters and multiple elements.
• Storyboarding and scene construction: Supports film storyboarding, wide-angle/low-angle shots, and other complex compositions.

Technical principles of HiDream-O1-Image-1.5

• Native omni-modal UiT architecture: The model is based on HiDream.ai's self-developed Unified Transformer (UiT) native omni-modal architecture. The architecture uses a unified pixel-level native representation to process multimodal information, avoiding the information loss caused by modality conversion in traditional multimodal models, enabling text, images, and other data to complete understanding and generation within a unified space.
• From open-source validation to commercial production: The model continues the technical路线 of the open-source version HiDream-O1-Image-Dev-2604, advancing the UiT architecture from technical validation to production validation. Based on the pixel-level native omni-modal capabilities already verified in the open-source version, the commercial version is enhanced and optimized for high-demand commercial scenarios such as advertising and marketing, brand design, and e-commerce visuals, achieving the transformation of underlying architectural advantages into a visual productivity tool.
• Comprehensive capability enhancement mechanism: By improving semantic adherence precision, stability in complex image generation, text rendering accuracy, and multi-subject consistency control, the model achieved an ELO of 1265 in an anonymous comparative evaluation with over 4,000 samples. The technical core lies in end-to-end joint modeling of deep semantic understanding of text instructions and pixel-level image generation, ensuring the coordinated unity of complex composition, spatial perspective, and visual storytelling.