Z-Image Power Nodes is a collection of nodes designed specifically for the Z-Image / Z-Image Turbo model. They are based on some ideas and discoveries I made while developing the Amazing Z-Image Workflow.
This page includes example workflows that utilize these nodes. The nodes can be installed via ComfyUI-Manager or manually from the GitHub repository. For more information, please visit:
These are the nodes currently included in the project:
⚡Z-Sampler Turbo ^G2
A specialized sampler designed for Z-Image Turbo that achieves sufficient quality to eliminate the need for further post-processing.⚡Style Prompt Encoder
Applies a selected visual styles to your prompt and encodes both of them using a text-encoder model (clip).⚡Style String Injector
Seamlessly integrates a chosen style into your prompt text. It accepts a string as input and modifies it based on the selected style.⚡My Top-10 Styles
Allows you to create a list of favorite styles for quick selection of your most used ones.⚡VAE Encode (for Soft Inpainting)
Encodes images into a latent representation, embedding the mask that indicates where inpainting will be applied.⚡Save Image
Saves generated images with the option to embed CivitAI-compatible metadata, making it easy to share generation parameters through that platform.⚡Empty Z-Image Latent Image
Creates an empty latent image of the appropriate size for Z-Image, selecting aspect ratio, scale, and orientation.
Description
The most important changes and new features in this release:
Additional Styles: Now featuring 70 pre-defined styles to choose from.
New "Style & Prompt Encoder" Node: Allows selecting ANY style and combining it with the prompt, generating embedings.
New "Style String Injector" Node: Directly injects style text into your prompt strings.
Enhanced "Save Image" Functionality: More reliable metadata export to CivitAI, including support for sub-graphs.
Console Information: Internal details such as the number of available styles are now displayed in the console.
Bug Fixes: Several issues have been resolved to improve overall stability.
Deprecated Nodes:
The "Illustration-Style Prompt Encoder" node has been deprecated.
Its functionality is now available in the new "Style & Prompt Encoder" node.The "Photo-Style Prompt Encoder" node has been deprecated as well.
Its features are integrated into the new "Style & Prompt Encoder" node.
FAQ
Comments (16)
The update is great, just for fun, I wish there was a "random mode" for each category :)
Yes!! I actually thought about that too! I'm pretty sure I'll get around to adding it eventually. It's something I'd really love to have. haha. Maybe in the next version.
and to add a little spice, random lora node/s (trying to do this now)
this actually amazing. Thank you for these nodes, really
You're welcome! Happy to hear they were helpful.
This works very, very well!
1. Most of the photo styles are pretty muted/flat in terms of color.
2. How do you add a custom style?
3. Any chance this can be combined with ControlNet?
1. "Yes, I've noticed that too. About two weeks ago, I came up with an idea to enhance color vibrancy and contrast. So in the upcoming v1.0 version, I included a new parameter inside the 'Z-Sampler Turbo' node called 'initial_noise_calibration', which should help improve this issue. I wrote something documenting this: https://github.com/martin-rizzo/ComfyUI-ZImagePowerNodes/blob/master/docs/zsampler_turbo.md
2. To add a custom style, connect a ComfyUI utils > primitive > String (Multiline) node to the 'customization' input of the 'Style & Prompt Encoder'. Write your style template in the String (Multiline) node. Currently, custom styles are defined within the specific workflow where this connection is made, rather than globally. However, this allows you to reuse workflows with different prompts while maintaining your custom style.
3. Regarding combining it with ControlNet... I doubt. My current priority is trying to secure a sponsor or patron to research more efficient ways to train LoRAs for Turbo.
Thank you for your interest in my nodes, and feel free to reach out if you have more questions.
This is absolutely incredible. Have some buzz! Thank you for sharing.
Thank you so much for your support! I'm glad to hear that you found these nodes useful. With the release of v1.0, I've included some significant improvements that may help you achieve even better image results. Best regards!
Would it be possible for you to create a new type of ksampler that would allow modifying the denoise to make images with double sampler? (similar to a hires-fix), it improves a lot when you use 2 samplers in z-image. I currently use your node to generate the image and a second ksampler with simple euler, and the images come out incredible, but I would like to use your sampler in both steps. Would that be very difficult?
In the upcoming v1.0 release, the 'Z-Sampler Turbo' node now includes a fully functional 'denoise' parameter. I had to implement some custom hacks to make it work, so there's no guarantee it will perform well in all situations. Once the new version is out, please give it a try in your workflow and see if you can find a denoise setting that works for you when using two samplers consecutively. I'd love to hear your feedback on how it performs.
I downloaded the custom_styles workflow from the github page but it is using 'StylePromptEncoder2' and 'ZSamplerTurbo2' nodes which I can't find in the custom node pack. Where would I find these?
Sorry for the confusion. At the time you wrote this message, I was about to release the new version but then came up with some additional ideas and postponed the launch. That's why those new nodes are missing. In version 1.0, everything should work properly now.
Also, the example workflows have been updated since you downloaded them.
This is outstanding work! Thank you so much for creating the nodes!
You're welcome! I'll keep working on improvements. Greetings
Please can you make a great workflow like this but for a really cool anime model called circlestone-labs/Anima. Great work as always.