Z Image Real Skin workflow | Realistic Portrait Generator
Creates portraits with real human skin texture and natural lighting.
AI model workflow
Z-Image ComfyUI Workflows
Create fast, photoreal images with ready-to-run Z-Image ComfyUI workflows for text-to-image, image-to-image, character editing, ControlNet, fine-tuned styles, and LoRA inference. Run them on RunComfy Cloud, download workflow JSON, or deploy supported workflows as APIs.
10 workflows
Featured workflow
Z-Image | Fast Photorealistic Base Model
The workflows
10 Z-Image ComfyUI Workflows
Each Z-Image ComfyUI example below is a real image workflow for Z-Image, Z-Image Turbo, Z-Image Base, Z-Image De-Turbo, Z-Image fine-tuned models, LoRA, ControlNet, image-to-image, character head swap, and photoreal generation. Run it on RunComfy Cloud, download the workflow JSON, or deploy supported workflows as APIs.
Z Image Turbo Headswap for Characters | Fast Swap Tool
Seamless head swap tool for perfect character customization.
Z-Image | Fast Photorealistic Base Model
Super-fast image maker with stunning clarity and total control.
Z-Image LoRA Inference | AI Toolkit ComfyUI
Apply your AI Toolkit-trained Z-Image LoRA in ComfyUI through a single RCZimage pipeline node for training-matched output.
Z Image ControlNet | Precision Image Generator
Total control over image poses, edges, and depth layouts.
Z Image Turbo | Ultra-Fast Photorealistic Generator
Generate ultra-clear visuals fast with unmatched real-time detail.
Z-Image I2I for Characters | Ultimate Photorealism
Turns portraits into lifelike, perfectly detailed realistic faces fast.
Z-Image Turbo LoRA Inference | AI Toolkit ComfyUI
Run your AI Toolkit-trained Z-Image Turbo LoRA in ComfyUI with training-matched defaults using a single RC custom node.
Z-Image Finetuned Models Collection | Multi-Style Generator
Create stunning, detailed images across multiple styles and moods easily.
Z-Image De-Turbo LoRA Inference | AI Toolkit ComfyUI
Run your AI Toolkit-trained Z-Image De-Turbo LoRA in ComfyUI with training-matched behavior using a single RCZimageDeturbo custom node.
10 of 10 workflows shown
About Z-Image
What is a Z-Image ComfyUI workflow?
A Z-Image ComfyUI workflow is an editable image node graph built around Alibaba's 6B-parameter Z-Image family. It connects a Z-Image checkpoint with prompts, text encoders, a VAE, sampling and decoding nodes, plus any reference image, ControlNet, LoRA, identity, or post-processing steps. A working graph lets you begin from a real result instead of assembling the model pipeline yourself.
Choose Z-Image Turbo when you want fast generation with a short sampling path, and choose Z-Image Base when you want the non-distilled foundation for compatible fine-tuning and LoRA workflows. Z-Image De-Turbo and fine-tuned variants provide alternative quality and style behavior, while image-to-image, head-swap, and ControlNet workflows add identity or structural guidance. Use the exact checkpoint and graph listed on the workflow page because these variants are not drop-in replacements for every setup.
On RunComfy, each Z-Image ComfyUI workflow page shows its real output, intended inputs, and required files. Click Run now to open the graph in a prepared cloud ComfyUI environment, or download the Z-Image workflow JSON to use locally. Your local installation still needs the matching model, text encoders, VAE, LoRAs, ControlNet weights, and custom nodes.
References and model resources: Z-Image GitHub, ComfyUI Z-Image Turbo tutorial, ComfyUI Z-Image examples, Explore Z-Image models on Hugging Face, Explore Z-Image models on Civitai.
Capabilities
What you can create with these Z-Image workflows
Z-Image ComfyUI workflows cover fast prompt generation, character-aware image editing, structural control, and custom LoRA or fine-tuned styles.
Fast text-to-image generation
Generate detailed, photoreal images from text prompts with a fast Z-Image workflow.
Image-to-image character editing
Retouch or transform a character image while retaining recognizable facial and identity details.
ControlNet-guided generation
Guide composition and structure with compatible control inputs instead of relying on the prompt alone.
LoRA and fine-tuned styles
Load a compatible Z-Image LoRA or fine-tuned checkpoint for trained characters, products, and visual styles.
Z-Image models on RunComfy
Choose the right Z-Image variant. Run the complete image graph.
RunComfy opens Z-Image workflows in native ComfyUI with the checkpoint, encoders, VAE, LoRAs, ControlNet weights, and custom nodes prepared. Add a prompt, reference image, or control input, then change resolution, sampling, guidance, identity strength, and post-processing in the graph. Use a cloud GPU when you want the full-quality pipeline without configuring or fitting it to local hardware.
Run a Z-Image workflow nowNative ComfyUI graphs for Z-Image generation and editing
Open Turbo, Base, image-to-image, head-swap, ControlNet, and LoRA pipelines as complete graphs with editable prompts and settings.
Z-Image checkpoints, LoRAs, and control weights in the cloud
Keep the compatible model, custom LoRAs, ControlNet weights, encoders, and supporting files with the workflow across sessions.
Custom nodes for identity and structure control
Use the nodes required for image-to-image retouching, head swaps, trained styles, structural guidance, and output enhancement.
Cloud GPU memory for full Z-Image graphs
Run higher-resolution, ControlNet, LoRA, and post-processing pipelines without removing nodes to fit a smaller local GPU.
Applications
What people use these Z-Image workflows for
Choose by the image result you need, from photoreal portraits and head swaps to multi-style outputs and rapid visual concepts.
Photoreal Portraits
Create natural-looking portraits with detailed skin texture, lighting, and facial features.
Character Head Swap
Replace a character's head or face while keeping the target pose, clothing, and surrounding image coherent.
Multi-Style Images
Compare fine-tuned Z-Image variants for realistic, illustrated, cinematic, or graphic visual directions.
Rapid Visual Concepts
Generate fast concept images for characters, products, scenes, and marketing ideas from a short prompt.
Frequently asked questions
Z-Image workflow FAQ
Answers for choosing Z-Image Turbo, Base, or Edit workflows, running GGUF or low-VRAM setups, rendering readable text, using LoRA or ControlNet, and downloading JSON.
What Z-Image ComfyUI workflow should I start with?
Start with a Z-Image Turbo text-to-image workflow when you want fast prompt-based image generation. Use an image-to-image or head-swap workflow when you already have a source image, and use a ControlNet or LoRA workflow when structure, identity, product style, or a trained visual look matters.
Should I use Z-Image Turbo, Base, or Edit?
Use Z-Image Turbo for the fastest finished image generation, Z-Image Base when you need the non-distilled foundation for fine-tuning or compatible LoRA work, and Z-Image Edit for instruction-following image editing. Pick the workflow built for that variant because Turbo, Base, Edit, and De-Turbo graphs are not always interchangeable.
Can Z-Image run on low VRAM with GGUF?
Yes, many ComfyUI users run Z-Image with GGUF or other quantized weights to reduce memory use. Low-VRAM setups can be slower, and changing prompts may require the text encoder to run again. Use a prepared Z-Image workflow when you want the right model loaders, encoder, VAE, and settings already wired.
Can Z-Image generate readable text in images?
Yes, Z-Image is known for stronger Chinese and English text rendering than many open image models, especially with clear prompts and enough resolution. Put the exact words in quotes, describe where the text appears, and keep the design simple when legibility matters.
Can I use Z-Image with LoRA and ControlNet?
Yes, when the LoRA or ControlNet model is made for the same Z-Image base used by the workflow. Use LoRA for a trained character, product, or style, and ControlNet for pose, edges, depth, or other structure. Do not assume FLUX or SDXL add-ons will work in a Z-Image graph.
Can I train my own Z-Image LoRA?
Yes. RunComfy Trainer supports Z-Image Turbo, Z-Image Base, Z-Image De-Turbo, and Z-Image L2P. Train against the same base you plan to use for inference.
Where can I download a Z-Image ComfyUI workflow JSON?
Click Download JSON on a Z-Image workflow card or detail page, then open the file in ComfyUI. A local run still needs the matching Z-Image checkpoint, text encoder, VAE, LoRA or ControlNet weights, and custom nodes. Click Run now to open the prepared workflow and run panel on RunComfy Cloud instead.
Related resources
More ways to use Z-Image
Train a Z-Image LoRA with the supported Turbo, Base, De-Turbo, or L2P base model. No matching Z-Image Models page is currently listed in RunComfy's model catalog.
Train a Z-Image Turbo LoRA
Train a fast custom character, product, or style LoRA using the supported Turbo training adapter.
Open Z-Image Turbo TrainerTrain a Z-Image Base LoRA
Use the non-distilled Z-Image base for a compatible custom LoRA training workflow.
Open Z-Image Base TrainerTrain a Z-Image De-Turbo LoRA
Train against the supported de-distilled Z-Image variant for matching inference workflows.
Open De-Turbo TrainerTrain a Z-Image L2P LoRA
Choose the L2P pixel-space base when your intended Z-Image workflow uses that architecture.
Open Z-Image L2P Trainer