Z-Image: Text to Image

Foundation for creative freedom. Diverse aesthetics with exceptional photorealistic quality; ideal for fine-tuning; responsive to negative prompts; high generation diversity.

Input

Describe the image.

Z-Image works well around 1-2 MP, e.g. 1344×768.

Z-Image works well around 1-2 MP, e.g. 1344×768.

Advanced settings

3-5 recommended.

Default: 4

What to keep out of the image.

Leave empty for a new result each run. Enter a number to repeat one.

30-50 recommended.

Default: 25

Output
Output
Example

试试一个示例

关于此工作流

The Z-Image: Text to Image workflow is designed to transform textual descriptions into high-quality images with a focus on photorealism and aesthetic diversity. Utilizing the Z-Image model, this workflow is particularly adept at producing images that are not only visually stunning but also highly responsive to both positive and negative prompts, allowing for fine-tuning and creative exploration.

The workflow employs specific nodes such as SaveImage for output management and MarkdownNote for documentation, alongside a unique node type identified by the ID 9b9009e4-2d3d-445f-9be5-6063f465757e, which plays a crucial role in the image generation process.

Technically, this workflow leverages the qwen_3_4b.safetensors text encoder and the z_image_bf16.safetensors diffusion model to interpret and render textual inputs into images. The process involves running the diffusion model for 30 to 50 steps with a classifier-free guidance (cfg) scale of 3 to 5, ensuring a balance between adhering to the prompt and introducing creative variability. This setup makes the workflow particularly useful for artists and designers seeking to generate diverse image outputs from textual descriptions, providing a foundation for creative freedom and exploration.

常见问答