Z-Image Turbo All-in-One ComfyUI Workflow
Generation, detailing, upscaling, finishing, ControlNet, inpainting and Flux Edit in one modular workflow.
This workflow is built as an all-in-one production pipeline for Z-Image Turbo inside ComfyUI .
The goal is simple: generate a strong base image, refine the important details, upscale it, polish the final look, and keep enough flexibility for pose control, image prompting, inpainting and perspective changes.
It is designed for people who want more than a simple text-to-image workflow. This is meant as a complete creative workspace for cinematic, photorealistic and stylized image creation.
What this workflow includes
The core generation is based on Z-Image Turbo , followed by several optional refinement stages:
Eye Detailer for sharper and cleaner eyes
Face Detailer for improved facial consistency and quality
SeedVR Upscaling for high-resolution output
Color Correction for final image balancing
HDR / contrast enhancement for a stronger cinematic look
Film Grain for a more organic final finish
LORA Loader
It also includes multiple control and utility sections:
ControlNet preprocessors
DWPose
Canny
Depth
etc...
Image-to-Prompt
Useful for analyzing reference images and building better prompts
Inpainting
For local fixes, corrections and creative edits
Flux Edit integration
Since Z-Image Turbo currently does not have a mature PuLID / IP-Adapter style identity workflow, Flux Edit is included as an additional editing option.
This can be useful for perspective changes, stronger image edits, pose adjustments or reworking an existing image while keeping the overall visual direction.
Why I made this
I wanted a workflow that feels closer to a full image production pipeline instead of a single generation graph.
Z-Image Turbo is fast and powerful, but for polished final images I often want additional steps: better eyes, better faces, cleaner upscale, stronger color grading, and tools for controlled composition.
This workflow combines those steps into one modular setup, so you can enable only the parts you need.
Recommended use cases
This workflow can be used for:
cinematic portraits
character images
fashion / lifestyle shots
realistic AI photography
social media visuals
controlled pose generations
image edits and variations
upscaled final renders
prompt extraction from reference images
inpainting and correction passes
Notes
You may need to install the required custom nodes and download the necessary models manually.
Model files are not included. Please check that you have the required Z-Image, SeedVR, ControlNet, detailer, inpainting and Flux Edit models installed in the correct ComfyUI folders.
The workflow is intentionally modular. You can bypass sections like upscaling, detailers, ControlNet or Flux Edit if you do not need them for a specific image.
Tested on RTX 4070 10GB and 5080 16GB
Main Models used:
z_ image_ turbo_bf16.safetensors
flux-2-klein-base-9b-fp8.safetensors
Qwen3-VL-2B-Instruct
sam3.safetensors
seedvr2_ema_3b_fp8_e4m3fn.safetensors
ae.safetensors
full_encoder_small_decoder.safetensors
qwen_3_8b_fp8mixed.safetensors
My goal
The goal of this workflow is to give creators a flexible starting point for a more complete Z-Image Turbo pipeline:
generate, control, repair, upscale and finish your image inside one clean ComfyUI workflow.
Feel free to modify it, improve it and adapt it to your own style.
If you create something cool with it, I would love to see your results!
