ComfyUI Node
Z-Image Generate (SDNQ)
A ComfyUI node in Z-Image (SDNQ) with 16 inputs and 1 output.
Z-Image Generate (SDNQ)
- pipeline
- input_image
- IMAGE
◄prompta beautiful landscape►
◄resolution_preset1024x1024 ( 1:1 )►
◄num_inference_steps9►
◄guidance_scale0.0►
◄seed42►
◄shift3.0►
◄max_sequence_length1024►
◄noise_scale1.00►
◄strength0.75►
◄custom_width1024►
◄custom_height1024►
◄enhance_promptfalse►
◄unload_after_generationfalse►
◄gc_cudafalse►
CategoryZ-Image (SDNQ)
Inputs (16)
| Name | Type | Default | Description |
|---|---|---|---|
| pipeline | ZIMAGE_SDNQ_PIPELINE | — | |
| prompt | STRING | a beautiful landscape | Text description of the image to generate. Be descriptive for best results |
| resolution_preset | COMBO | 1024x1024 ( 1:1 ) | Select a resolution preset, 'use image size' to match input image dimensions, or 'custom' to set manual width/height |
| num_inference_steps | INT | 91–100 | Number of denoising steps. Z-Image Turbo is optimized for 5-9 steps. More steps = slower but potentially higher quality |
| guidance_scale | FLOAT | 0.00–20 | Classifier-free guidance scale. Z-Image Turbo is trained for guidance_scale=0.0 (recommended). Higher values increase prompt adherence but may reduce quality |
| seed | INT | 420–18446744073709550000 | Random seed for reproducible generation. Same seed + settings = same image |
| shift | FLOAT | 3.01–10 | Time shift for FlowMatch scheduler. Controls sampling distribution: lower (1-2) = more noise-end sampling, higher (4-10) = more clean-end sampling. Default 3.0 is balanced |
| max_sequence_length | INT | 1024128–2048 | Maximum prompt token length. 1024 recommended for detailed prompts (Z-Image works best with long descriptions). 512 for faster speed |
| noise_scale | FLOAT | 1.000–2 | Scale for initial noise. 1.0 = normal random noise, lower = less variation, higher = more variation. Only affects pure noise (not image latents) |
| strength | FLOAT | 0.750–1 | Denoising strength for img2img. 0.0 = no change (pure image), 1.0 = full regeneration (ignore image). Only applies when input_image is provided. Controls how many timesteps to denoise |
| input_imageopt | IMAGE | Optional input image to encode to latents. Will be blended with noise based on noise_scale. Use 'use image size' preset to match image dimensions | |
| custom_widthopt | INT | 1024256–2048 | Custom output width (only used when resolution_preset is 'custom'). Must be divisible by 64 |
| custom_heightopt | INT | 1024256–2048 | Custom output height (only used when resolution_preset is 'custom'). Must be divisible by 64 |
| enhance_promptopt | BOOLEAN | false | Use Z-Image's built-in prompt enhancer for improved semantic understanding and detail. May not be available in all models |
| unload_after_generationopt | BOOLEAN | false | Move models to CPU after generation to free VRAM. Useful for memory-constrained setups |
| gc_cudaopt | BOOLEAN | false | Run Python garbage collection and clear CUDA cache after generation. Helps reclaim memory |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |