Summary This repository offers custom ComfyUI nodes tailored for Z-Image and Z-Image Turbo models. It targets ComfyUI users seeking enhanced control over image generation, enabling streamlined workflows, consistent artistic styles, and rapid iteration with minimal steps. The nodes aim to maximize Z-Image model capabilities through simplified prompting and novel control parameters.

How It Works Key nodes include the "Style & Prompt Encoder," which integrates over 100 predefined visual styles via a searchable gallery, automatically adjusting prompts for consistent artistic direction while preserving subject/composition. The "Z-Sampler Turbo" node ensures high consistency from 3+ steps, yielding acceptable results by 5 and high quality by 7. It features "Intensity" for contrast/saturation and "Intensity Bias" for noise calibration. "Turbo Creativity" uses latent scrambling for compositional variety without altering style/prompt, with optional "refined" modes for coherence at increased generation time. Utility nodes for VAE encoding, quick style selection, and CivitAI metadata embedding are also provided.

Quick Start & Requirements Installation is recommended via ComfyUI Manager or manual cloning into ComfyUI/custom_nodes. A recent ComfyUI version is required. Recommended checkpoints include GGUF (Q8/Q5), FP8, and BF16 versions of Z-Image Turbo and Qwen3-4B models. Specific safetensors and GGUF files for diffusion models, text encoders, and VAEs are listed. Example workflows are available in the /workflows directory.

Highlighted Details

• "Style & Prompt Encoder" with a searchable gallery of 100+ predefined styles.
• "Z-Sampler Turbo" maintains consistency from 3 steps, high quality by 5-7 steps.
• "Intensity" and "Intensity Bias" parameters for fine-tuning contrast, saturation, and noise.
• "Turbo Creativity" enhances compositional variety via latent scrambling.
• "Save Image" node embeds CivitAI-compatible metadata.

Maintenance & Community The project encourages community support via GitHub stars and Ko-fi. Specific links to community channels, roadmaps, or notable contributors are not detailed in the README.

Licensing & Compatibility Licensed under the MIT license, permitting commercial use and integration into closed-source projects.

Limitations & Caveats "Turbo Creativity" may introduce hallucinations; refined options increase generation time. FP8 checkpoints require careful testing due to potential precision loss from naive truncation. Parameter effectiveness (e.g., "Intensity Bias") depends heavily on prompt and style.