martin-rizzo/ComfyUI-ZImagePowerNodes

What it solves

This project provides a set of custom nodes for ComfyUI specifically designed to push the capabilities of the Z-Image / Z-Image Turbo model. It simplifies the process of achieving high-quality, consistent image generation with fewer sampling steps and provides tools for artistic control over style, contrast, and composition.

How it works

The toolkit consists of several specialized nodes:

• Z-Sampler Turbo: A custom sampler that maintains image consistency from as few as 3 steps. It includes an "Intensity" parameter to control contrast and saturation, and an "Intensity Bias" for brightness/focus calibration. It also features "Turbo Creativity" using latent scrambling to increase compositional variety without altering the visual style.
• Style & Prompt Encoder: A library of over 100 predefined visual styles with a searchable gallery, allowing users to shift artistic direction without complex prompting.
• Utility Nodes: Includes a VAE encoder for soft inpainting, a "Top-10 Styles" quick-selection node, and a save image node that embeds CivitAI-compatible metadata.

Who it’s for

Digital artists and ComfyUI users who are using the Z-Image / Z-Image Turbo models and want more precise control over their output, faster iteration times, and a streamlined style management system.

Highlights

• Low-Step Consistency: High-quality results possible with 7 or more steps, with acceptable results from 5 steps.
• Style Library: Over 100 predefined styles with thumbnail previews for easy artistic direction.
• Compositional Variety: "Turbo Creativity" increases variety in posing and framing while keeping colors and style consistent.
• CivitAI Integration: Ability to save images with metadata that is easily shareable on CivitAI.