Recursos

O Z-Image-Omni-Base marca uma mudança estratégica do modelo 'Base' original para uma arquitetura de pré-treinamento 'omni' (onipotente). Ele unifica as tarefas de geração de imagem e edição/inpainting dentro de uma única estrutura usando o Scalable Single-Stream Diffusion Transformer (S3-DiT). Este pré-treinamento omni permite transições perfeitas entre a geração de novas imagens e a edição das existentes sem a necessidade de modelos especializados separados, oferecendo maior eficiência de parâmetros e flexibilidade para desenvolvedores.

O Z-Image é um modelo de fundação de geração de imagem poderoso e altamente eficiente com 6B parâmetros. Aproveitando a arquitetura Scalable Single-Stream Diffusion Transformer (S3-DiT), ele processa texto, tokens semânticos visuais e tokens VAE de imagem como um fluxo unificado. O Z-Image serve como núcleo para variantes como Z-Image-Turbo e Z-Image-Omni-Base, oferecendo desempenho de ponta entre modelos de código aberto.

Z-Image is a powerful and efficient image generation model with 6B parameters. Currently there are three variants: Z-Image-Turbo (distilled version, only 8-step inference), Z-Image-Omni-Base (base model) and Z-Image-Edit (image editing variant).

Experience Z-Image Turbo model on ModelScope platform, providing online inference service and API interface for quick integration and use by developers.

Z-Image Turbo model in HuggingFace community, providing complete model weights, usage examples and community support. Monthly downloads reach 111,244.

Z-Image De-Turbo is a de-distilled version of Tongyi-MAI/Z-Image-Turbo, fine-tuned on images generated by Z-Image-Turbo to break down the turbo distillation limitations. This model is specifically designed for training and deep fine-tuning, offering enhanced trainability and flexibility compared to the original turbo model.

Red-Z-Image-AIO-1.5 is a community-enhanced version of the official Z-Image-Turbo model, built on the S³-DiT single-stream diffusion architecture. It addresses key pain points like setup complexity, performance on low-end hardware, and creator-specific needs, including specialized NSFW tuning for realistic rendering of human anatomy.

Official ComfyUI documentation for Z-Image Turbo, featuring complete workflow templates, detailed model download instructions, and optimization settings for low VRAM devices. Perfect for users who want local deployment with full control over the generation process.

PrunaAI's optimized version of Tongyi-MAI's Z-Image Turbo, accelerated through advanced compression techniques. This version applies smart caching, model compilation, and quantization to make image generation even faster while preserving the original model's photorealistic quality and excellent Chinese text rendering capabilities.

GGUF-Org provides officially converted GGUF quantized versions of Z-Image Turbo, enabling deployment on consumer-grade GPUs with as low as 6GB VRAM. Supports multiple quantization levels including Q3_K_S, IQ4_NL, and IQ4_XS for different VRAM/quality trade-offs.

Qwen3-4B-GGUF is the required text encoder for Z-Image GGUF deployments, providing bilingual (Chinese/English) understanding and advanced reasoning capabilities. Supports unique thinking mode for complex logical reasoning tasks.

Community-maintained GGUF model collection by Jayn7, providing multiple quantization variants of Z-Image Turbo with detailed ComfyUI setup guides. Popular choice with over 200 likes in the community.

This LoRA model enhances Z-Image's existing pixel art capabilities, making them more detailed and refined. No trigger words required, but using "pixel art" in prompts can achieve better results.

This is a ControlNet model with 6 blocks added, trained from scratch on 1 million high-quality image dataset for 10,000 steps, supporting multiple control conditions.

Qwen-Image-i2L is an innovative model that takes images as input and directly outputs LoRA weights trained on those images. The system includes four specialized models: i2L-Style for style transfer, i2L-Coarse and i2L-Fine for content preservation, and i2L-Bias for output alignment with Qwen-Image aesthetics.

Z-Image Turbo official online demo application, maintained by Tongyi-MAI team, providing a platform to directly experience Z-Image generation capabilities. Runs on Zero GPU, no local configuration required.

Z-Image gallery application on ModelScope platform, showcasing Z-Image model generation effects in gallery format, providing Chinese interface optimization and integrated ModelScope ecosystem.

Smart frame application based on Z-Image, specially developed for MCP 1st birthday commemorative project, showcasing Z-Image's possibilities in creative application scenarios.

Z-Image custom LoRA model online gallery, hosting multiple custom-trained LoRA models including radical art styles and identity models, supporting online switching and preview functionality.

Interactive mystery game application developed based on Z-Image, combining image generation and gameplay, showcasing Z-Image's innovative use in entertainment application scenarios.

This paper proposes the DMDR framework, integrating reinforcement learning techniques into the distribution matching distillation process. Research shows that for reinforcement learning of few-step generators, the DMD loss itself is more effective than traditional regularization methods.

Z-Image project official homepage, providing comprehensive project introduction including core features, model architecture, performance evaluation and technical details. Supports bilingual content in Chinese and English.