The commit introduces two workflows in the `/workflows/experimental` directory:
"ZIT_experimental_styletester.json" (formerly `style_tester.json`) and the new
"ZIT_experimental_text2image.json". This latter workflow provides a basic
text-to-image functionality using the newly implemented experimental nodes.
The commit enhances the `style_tester.json` workflow by introducing configurability
for the number of sub-images to generate, ranging from 4 to 8. This improvement
addresses the previous limitation where 8 images were always generated, even if
fewer styles were being tested, thus optimizing the generation process and reducing
unnecessary wait times.
The commit refines the `style_tester.json` workflow to accommodate testing
with eight different visual styles using the same prompt. Previously, the
workflow supported only six styles.
The commit adds a `README.md` file to the `/workflows` directory,
explaining the purpose of the "/workflows/experimental" and
"/workflows/release" subdirectories. Additionally, it updates links
in the main project `README.md` to point directly to the released
workflows in the "/workflows/release" folder.
The commit reorganizes the "/workflows" directory by dividing it into
`/workflows/release` and `/workflows/experimental` folders. This
structure helps in categorizing workflows, with the "release" folder
containing tested and finalized workflows to be included in published
versions, while "experimental" houses experimental workflows that are
still under testing and development.
This commit introduces a new "Z-Sampler Turbo ^G2 (Simple)" node that exposes
users to fewer parameters, mostly boolean switches, while maintaining the full
functionality of zsampler_turbo_core(..) method. By limiting parameter ranges
to those with significant impact on the final image and retaining only influential
settings, this node simplifies the sampling process. It allows for quicker
experimentation with specific prompts to find optimal configurations efficiently.
The existing "Z-Sampler Turbo ^G2" node has been renamed to "Z-Sampler Turbo ^G2 (Extended)"
to clearly differentiate between the simpler and more extensive versions, aiding user
understanding and selection.
This commit relocates the comparative figures from the main README,
which illustrate how each parameter affects the final image, to the
documentation directory under `docs/figures`. Additionally, it
includes the corresponding workflow `.json` files that generated
these figures.
This commit updates the README file in the "/workflows" directory to
document three previously undocumented example workflows. The update
provides a brief description for each workflow, enabling users to
understand their purpose and usage more easily.
This commit modifies the default style in the reference text2image
workflow to "High Key Cinematic Portrait", providing users with a
different artistic starting point for their image generation.
This commit improves compatibility with CivitAI by adding metadata extraction
from the new "Z-Sampler Turbo ^G2" node and enabling the retrieval of prompts
from "String (Multiline)" nodes tagged with ">>C".
The workflow "z-image-turbo__t2i_double_trouble.json" has been updated to
utilize these enhancements, generating images that CivitAI can automatically
extract information from, including prompts.
This commit updates the "/workflows" directory by adding FP8 and BF16 versions
of previously created GGUF workflows. These new format-specific workflows are
automatically generated using the `convert.sh` script and stored in their
respective directories: "/workflows/FP8" and "/workflows/BF16".
This commit renames the "/workflows/__zdev" directory to "/workflows/_z_".
The new name is intended to maintain a low profile for this development-specific
directory, where experimental and internal workflows are stored.
This commit enhances the "/workflows/__zdev" directory by adding image
files (in JPG format) and their corresponding ComfyUI workflows used to
generate the figures in the README.
These figures illustrate how each user-accessible parameter ('intensity',
'intensity_bias', 'steps', 'styles', and 'turbo_creativity') affects the
final image output, providing a visual guide for users to understand and
utilize these parameters effectively.
This commit refines the organization of the "/workflows" directory by
removing the subdirectory containing outdated workflows from previous
versions. Additionally, a new subdirectory named "__zdev" is established
to house workflows created for development purposes.
This commit introduces three core reference workflows to guide users in
performing key tasks: `z-image-turbo__text2image`, `__t2i_double_trouble`,
and `__inpainting`.
These workflows serve as starting points, providing a solid foundation for
users to build upon and customize according to their specific needs.
This commit introduces three additional workflows that serve as examples of
how to use various nodes within the project. The new workflows are:
- forging_custom_styles: Demonstrates how to create and customize styles.
- injecting_style_into_string: Shows how to inject style into a prompt string.
- selecting_my_top_10_styles: Illustrates using the "My Top-10 Styles" node.
These examples aim to provide users with practical insights into leveraging
different nodes effectively.
This commit introduces three new workflows used to generate the figures
in the documentation. These workflows showcase how different parameters
(style, intensity, and 'turbo creativity') affect the final image results.
This commit updates the "Z-Sampler Turbo" nodes to use a set of more intuitive
parameters: 'intensity', 'intensity_bias', 'initial_sample_size', 'turbo_creativity',
and 'turbo_creativity_extra_steps'.
These values are internally converted within the nodes to match the requirements
of the `zsampler_turbo_core` method. The goal is to present users with easily
understandable parameters that can be directly linked to the final image results,
while abstracting away the implementation details of `zsampler_turbo_core(..)`.
This commit introduces a new workflow example, "custom_styles.json",
demonstrating how to define and use custom styles within the project.
Notably, while similar workflows existed in previous versions, this
new example utilizes the Z-Sampler Turbo Gen2, which will be included
in the upcoming version 2.0 release.
This commit introduces two new Figures to showcase the impact of
different styles on image generation using the same prompt, and
another Figure illustrating how enabling "turbo creativity" mode
increases variation in image compositions.
Additionally, the existing "intensity" Figure has been refined by
adding a footer with the prompt text used for generating the images.
This commit introduces a new Figure demonstrating how the 'intensity'
parameter affects image generation, showcasing five images created with
the same prompt but varying intensity values. The Figure is added to the
repository along with the ComfyUI workflow used to generate it, providing
users with a clear visual understanding of the impact of different intensity
settings on output images.
This commit refines the "Requirements" and "Checkpoint Files" sections in the README.md file within the "/workflows" directory. The updates aim to maintain clarity while minimizing verbosity.
This commit refines the opening paragraph of the "Checkpoint Files" section in the README.md file within the "/workflows" directory. The update simplifies the text and gets straight to the point, improving clarity for users.
This commit updates the README.md file within the "/workflows" directory to reflect the new directory structure and document the use of different checkpoint formats (BF16, FP8, GGUF) across the three workflow groups. The documentation provides clearer guidance on how workflows are organized.
This commit reorganizes the "/workflows" directory in preparation for
the upcoming version 2.0 release. Previous workflow files have been
temporarily moved to a subdirectory named "__old" and will be removed
after the transition period.
The new directory now includes three subdirs ("BF16", "FP8", and "GGUF")
grouping workflows by checkpoint type. Workflows are created in "/GGUF"
and automatically converted to other formats (BF16 and FP8) using a script
called "convert.sh".
Additionally, the "/scripts/wfmigrator.py" script has been implemented to
handle these conversion operations internally.
This commit enhances the Z-Sampler Turbo Gen2 nodes by introducing two
optional inputs for connecting different conditioning to stages 2 and 3
of the sampler. These inputs allow for stage-specific conditioning,
while leaving them disconnected defaults to using the main positive
conditioning input for all stages.
This commit involves a comprehensive review of all example workflows located
in the `/workflows` directory. Errors have been corrected, notes added for
clarity, and some parameters adjusted to ensure each workflow functions
optimally and is easy to understand. Additionally, new workflows have been
created that utilize safetensors format checkpoints, duplicating existing
ones but modifying model loading nodes to support safetensor file types.
This commit updates five example workflows located in the /workflows directory.
The changes include adapting to recent node modifications, adding README notes
for usage instructions and parameter explanations, and implementing specific
improvements.
The updated workflows are:
- advanced__develop_and_test_new_styles.json
- advanced__zsampler_turbo_advanced.json
- custom_styles.json
- inject_style_into_string.json
- inpainting.json.
This commit completes the implementation of inpainting functionality
using the "Z-Sampler Turbo" node in conjunction with the newly added
"VAE Encode (for Soft Inpainting)" node. A ready-to-use workflow,
named "inpainting.json," has been configured and placed in the
"/workflows" directory for users to leverage these nodes effectively.
This commit updates the filename of `z-image_turbo_power_workflow.json` to
`z-image_turbo_text2image_workflow.json` within the "/workflows" directory.
The new name more accurately describes the functionality of the workflow,
enhancing clarity for users.
This commit enhances the workflows/README.md file by listing each available
workflow within the directory and providing a brief description for each.
These descriptions aim to guide users in understanding the purpose of each
workflow, improving overall usability.
This commit enhances the organization within the "/workflows" directory
by renaming some files to more descriptive names and flattening the
structure to reduce nested subdirectories, making it easier for users
to navigate the example workflows.
This commit rectifies an incorrectly placed link to the "ComfyUI-GGUF"
nodes within the documentation, ensuring accurate reference for users
interested in integrating these specific node functionalities.
This commit improves the "Requirements" section of the "/workflows/README.md"
file by providing a more detailed and comprehensive overview of necessary
prerequisites. It explains that while general requirements are covered, custom
modifications made by users may not be guaranteed to function perfectly due to
potential variations in implementation. Installation instructions for each
requirement are not included here, as dedicated sections elsewhere in the README
handle these details.
This commit begins the process of improving organization within the "/workflows" directory by restructuring and enhancing its README.md file. The aim is to provide clearer and more structured workflow documentation.
This commit enhances the development workflow used for evaluating
new styles ("evaluate_style.json") to accommodate 16 slots for
potential prompts. New prompts have been added, allowing users to
selectively activate or deactivate individual prompts during
testing to assess their performance with the given styles effectively.
This commit updates the internal development workflow "evaluate_style.json"
to include the new "initial_noise_calibration" option set at 100% for all
test images. This adjustment ensures that evaluations of new style templates
across multiple prompts consider the impact of this newly activated parameter.
This commit organizes the sampler nodes by removing the deprecated
status from "Z-Sampler Turbo" node, and renaming the class name of
the advanced node from ZSamplerTurboAdvanced2 to ZSamplerTurboAdvanced.
These changes reflect the consolidation of these two main nodes,
eliminating redundant variations and simplifying user access
to easy configuration settings for both basic and advanced nodes.
This commit focuses on streamlining the user interface for sampler options
to ensure they are as intuitive and accessible as possible. Two nodes,
"Z-Sampler Turbo (Advanced)" and the original "Z-Sampler Turbo," are
retained. The latter now includes new controls for initial noise, enabling
users to achieve images with enhanced contrast and saturation. Both nodes
offer adjustable levels for this initial noise correction.
Additionally, the temporary "Z-Sampler Turbo (2)" (ZSamplerTurbo2) node is
removed as its functionality is fully covered by the updated configurations
of the two main nodes.
This commit updates the `custom_styles.json` workflow to incorporate
recent modifications in the nodes. This workflow enables users to
configure their own custom styles.
This commit updates the primary example workflow "z-image_turbo_main_workflow"
to incorporate recent enhancements and new functionalities added to the nodes.
This ensures that the workflow remains current and benefits from the latest
improvements.
This commit includes an example workflow for mask inpainting, located in the
'workflows/inpainting_example.json' file. The addition leverages the recent
enhancements to the "Z-Sampler Turbo" node, which now supports mask inpainting
and adjustable denoising levels, providing users with a practical demonstration
of these features.
This commit standardizes parameter and variable names across the
"Z-Sampler Turbo" and "Z-Sampler Turbo (Advanced)" nodes. For example,
in the advanced node, parameters previously named 'offset' are now
consistently referred to as 'bias', ensuring uniformity and clarity
throughout both implementations.
This commit introduces the new "Z-Sampler Turbo (Advanced)" node,
an enhanced version offering improved functionality compared to the
existing sampler. The advanced sampler includes adjustable initial
noise parameters, which may address previous issues of low contrast
and saturation. Although still under testing, preliminary results
indicate significant improvement in performance.
This commit revises three example workflows, originally designed for older
versions, to incorporate the latest nodes that feature style selection through
the style gallery, showcasing the enhanced functionality in practical scenarios.
This commit significantly improves the code quality of the ComfyUI frontend
extension related to the "My Top-10 Styles" listing and editing nodes. Changes
include adding documentation, removing dead code, and moving helper functions
to common.js, resulting in a more maintainable and cleaner codebase.
The node displaying the user's top 10 styles is now named "My Top-10 Styles",
while the node allowing users to edit their preferred styles is titled
"My Top-10 Style Editor". These changes aim to make the functionality more
intuitive and easier for users to understand within ComfyUI's interface.
This commit represents a preliminary success in automatically renaming the
widgets that display each style when the user modifies them through the
configuration node. While full compatibility with Nodes 2.0 has not been
achieved yet, there has been significant progress towards ensuring dynamic
and responsive updates to the UI elements based on user changes.
Martin Rizzo