Z-Image All-in-One Creator
(Txt2Img / Img2Img / Inpaint / Upscale / ControlNet / Variation / Face Detailer)
I wouldn't say no to a cup of tea ;) : https://ko-fi.com/light_x02
This workflow is built around Z-Image Turbo and designed as a modular, extensible generation hub.
You get multiple generation modes, flexible sigma methods, advanced upscaling, rich prompt sources and automatic metadata saving, all wired to stay readable and easy to extend.
✨ Main Features
Multiple generation modes
Text-to-Image (Txt2Img)
Img2Img
Inpaint
Inpaint with SAM2 Segmentation (automatic region selection, e.g. face / person / background)
Prompt sources
Manual prompts (Positive / Negative)
Local Media Manager (reuse prompts from your own images with metadata)
Civitai Discovery Hub (import prompt, negative, metadata + image from Civitai)
QwenVL (generate prompts from an input image using an LLM)
Sigma methods
Standard method (classic behavior)
Method Sigma Low Noise – softer result, controlled denoising, still detailed
Method Sigma High Noise – stronger, punchier, higher energy with more visible noise
Noise & variation tools
SeedVarianceEnhancer to increase seed diversity for Z-Image Turbo
Multiple sampler/scheduler combos tuned for Z-Image (Euler, dpmpp, seeds_2 / seeds_3, res, etc.)
ControlNet support
Depth / DW / Canny / HED preprocessing
Control image megapixel control and strength sliders
Face Detailer
Acts like a refiner only on the face, especially useful for small or distant faces
Separate steps, CFG, sampler, scheduler and denoise controls
Upscale options
No Upscale (direct output)
Tile Upscale using UltimateSDUpscale
SeedVR2 Upscale for very small sources → high resolution output
Save & metadata
Multiple save points (1st pass / Upscale / Final)
Recommended: Save Image Final (with filters applied)
Automatic ComfyUI metadata extension for Civitai-compatible workflows
🔔 Note: Img2Img and Inpaint are NOT compatible with Method: Sigma High Noise.
They work normally with other methods (Standard, Sigma Low Noise, etc.).
🧩 Modular Design
The workflow is built to be easily extendable.
You can add:
new generation modes
new generation methods
new upscale methods
new prompt sources
By following a simple convention:
Generation modes
Add a group inside the
Generation Previewsubgraph.The group name must contain
generationand be cyan.The Generation Mode node will detect it and allow enabling/disabling modes.
Connect the outputs to the appropriate switches.
Generation methods
Add a group inside
Generation Preview.The group name must contain
method.Connect all inputs (switches / pipes) and route the latent output to
Latent Switch 2.
Upscale methods
Add a group inside the
Upscalesubgraph.The group name must contain
Upscaleand be green.Connect the inputs from the pipe and route the image output to
Image Switch Upscale.
Prompt sources
Connect the text output of your new source to
Prompt P Switch(in theBackendgroup on the right).Also connect your node to
Source Prompt, so the system can automatically disable it when another prompt source is selected.
🧠 Recommended Usage & UI Tips
This workflow is designed to be used with node link visibility turned OFF in ComfyUI:
cleaner visual layout
better performance on large workflows
Enable link display mainly when you want to modify or extend the workflow.
For best quality:
Use Method Sigma Low Noise for cleaner, softer images with strong detail.
Use Method Sigma High Noise when you want more punch, sharper noise and aggressive detail.
Upscale:
For typical stills, SeedVR2 around ~6MP target often gives very strong results.
For simple upscale, the Tile Upscale branch remains a good, classic option.
📦 Required Custom Nodes
ComfyUI Manager should normally detect and install missing nodes automatically.
Here is the full list of required extensions for this workflow:
ComfyUI-FRED-Nodes_v2
Comfy-image-saver
https://github.com/alexopus/ComfyUI-Image-Savercomfyui_controlnet_aux
https://github.com/Fannovel16/comfyui_controlnet_auxComfyUI-Impact-Pack
https://github.com/ltdrdata/ComfyUI-Impact-PackComfyUI-Custom-Scripts
https://github.com/pythongosssss/ComfyUI-Custom-ScriptsComfyUI_LayerStyle
https://github.com/chflame163/ComfyUI_LayerStylergthree-comfy
https://github.com/rgthree/rgthree-comfyComfyUI-Easy-Use
https://github.com/yolain/ComfyUI-Easy-UseComfyUI-RMBG
https://github.com/1038lab/ComfyUI-RMBGComfyUI-SeedVR2_VideoUpscaler
https://github.com/numz/ComfyUI-SeedVR2_VideoUpscalerComfyUI_UltimateSDUpscale
https://github.com/ssitu/ComfyUI_UltimateSDUpscaleComfyUI_essentials
https://github.com/cubiq/ComfyUI_essentialsComfyUI-Detail-Daemon
https://github.com/Jonseed/ComfyUI-Detail-DaemonComfyUI-QwenVL
https://github.com/1038lab/ComfyUI-QwenVLComfyUI-Impact-Subpack
https://github.com/ltdrdata/ComfyUI-Impact-SubpackComfyUI-Allor
https://github.com/Nourepide/ComfyUI-AllorComfyUI-mxToolkit
https://github.com/Smirnov75/ComfyUI-mxToolkitcomfyui_image_metadata_extension
https://github.com/edelvarden/comfyui_image_metadata_extensionSeedVarianceEnhancer
https://github.com/ChangeTheConstants/SeedVarianceEnhancerComfyUI-Civitai-Discovery-Hub
https://github.com/Light-x02/ComfyUI-Civitai-Discovery-HubComfyUI-Lightx02-Nodes
https://github.com/Light-x02/ComfyUI-Lightx02-NodesComfyUI-checkpoint-Discovery-Hub
https://github.com/Light-x02/ComfyUI-checkpoint-Discovery-HubComfyUI_Local_Media_Manager
https://github.com/Firetheft/ComfyUI_Local_Media_ManagerComfyUI-RvTools_v2
https://github.com/r-vage/ComfyUI-RvTools_v2ComfyUI-Flux-Continuum
https://github.com/robertvoy/ComfyUI-Flux-ContinuumComfyUI-KJNodes
https://github.com/kijai/ComfyUI-KJNodes
❤️ Credits & Support
This workflow was created by Light_x02:
https://civarchive.com/user/Light_x02
It required many days and countless hours of work to design, refine and document.
If you enjoy using this workflow, feel free to leave a comment (positive or constructive) on Civitai.
If you’d like to support the project, you can also offer a coffee on Ko-fi: