skills.video logo
开始使用
Z.ai提供方概览

Z-Image 图片 生成器

Z-Image 提供聚焦型 AI 图片生成能力，适合快速迭代、细节清晰和灵活视觉风格的创作。你可以在下方探索可用模型并直接开始生成。

探索 Z-Image 的模型

直接进入你想比较、测试或用于生成的具体模型页面。

Z-Image

Z-Image 的功能亮点

汇总该提供方主要模型系列中的共性优势。

High-Quality Creative GenerationStrong Diversity And ControllabilityWide Style Range And Negative PromptingFlexible Resolution And Sampling Controls

Z-Image

High-Quality Creative Generation

The official repository describes Z-Image as the foundation model behind Z-Image-Turbo, focused on high-quality generation and stronger aesthetic output. It is positioned for creators and developers who need a more capable base model for image synthesis rather than only the fastest distilled path.

查看 Z-Image
High-Quality Creative Generation

Z-Image

Strong Diversity And Controllability

The README explicitly states that Z-Image focuses on strong diversity and controllability. It highlights diversity across identities, poses, compositions, and layouts, which makes the model more suitable for iterative creative exploration and downstream customization.

查看 Z-Image
Strong Diversity And Controllability

Z-Image

Wide Style Range And Negative Prompting

The official repo says Z-Image supports a wide range of artistic styles and effective negative prompting. In the recommended parameters for the base model, negative prompts are described as strongly recommended for better control when removing unwanted content or tightening the composition.

查看 Z-Image
Wide Style Range And Negative Prompting

Z-Image

Flexible Resolution And Sampling Controls

For the base Z-Image model, the repository recommends resolutions from 512×512 up to 2048×2048 in total pixel area at any aspect ratio. It also recommends guidance scale 3.0 to 5.0, inference steps 28 to 50, and using cfg_normalization false for general stylism or true for realism.

如何在 skills.video 中使用 Z-Image

01

选择一个 Z-Image 模型

先挑选最适合你工作流的模型系列，再进入对应的模型详情页。

02

查看示例与功能

在生成前先检查每个模型的示例、控制项和功能细节。

03

开始创作

选好模型后，直接进入图片或视频生成流程。

图片模型

对比 Z-Image 的图片模型，适用于文生图与图像编辑工作流。

1 个模型
Z.ai

Z-Image

Z-Image

Z-Image 图像生成模型

常见问题

关于 Z-Image 模型和工作流的常见问题。

What is Z-Image?
Z-Image is the foundation generation model in Tongyi-MAI's Z-Image family. The official repository describes it as a 6B-parameter image generation model focused on high-quality output, richer aesthetics, diversity, and controllability.
How is Z-Image different from Z-Image-Turbo?
According to the repository, Z-Image is the higher-step foundation model, while Z-Image-Turbo is the distilled fast variant. Z-Image-Turbo targets very low-latency generation, whereas Z-Image is positioned more for higher-quality generation, creative development, and fine-tuning.
What resolutions does Z-Image support?
The official recommended range for the base Z-Image model is 512×512 to 2048×2048 total pixel area, with any aspect ratio.
What generation settings are recommended for Z-Image?
The repository recommends guidance scale 3.0 to 5.0, inference steps 28 to 50, and strongly recommends negative prompts for better control. It also notes that cfg_normalization false is suited to general stylism, while true is better for realism.
Does Z-Image support negative prompts?
Yes. The official README explicitly says negative prompts are strongly recommended for the base Z-Image model when you want better control over unwanted content.
What architecture does Z-Image use?
The project states that Z-Image uses a Scalable Single-Stream DiT (S3-DiT) architecture that combines text, visual semantic tokens, and image VAE tokens into one unified input stream.