MasterPrompt - One prompt to rule them all! + Wildcards版本Legacy v1.0 (Free WCards) (Free WCards)

I'm extremelly excited to share with you my Unprompted wildcards AND our powerful MasterPrompt project with a FREE version of the program.

In the downloads there are included the WILDCARDS from the Legacy 2.1.0 version of the program AND the 2.1.0 VERSION logic files!! (They require UNPROMPTED EXTENSION, it's not the same as the common Wildcards extension) (Check the 1.0 Legacy version images to learn how to use them once you donwload them from the 5.0.0 version page; to install the 2.1.0 wildcards drop the folders inside "unprompted" extension folder).

The 5.0.0 version of the program, with it's way more advanced logic files and wildcards shown in the promo video, are supporters content, go grab it from our Ko-Fi support page! This and many more advantages and surprises are waiting for those who join the crew!

https://ko-fi.com/post/MasterPrompt-5-0--Feral-Instinct-Y8Y0W38XT

5.0.0 is able to generate anything from jawbreaking photorealism and slice of life scenes to fantasy and extremelly aesthetic digital or anime art. It can generate males and females, crowds of people, animals, plants, humanoid hybrids, fantasy creatures, landscape, interior and exterior architecture. From almost any angle and distance. When unbounded and combined with high quality checkpoints it reaches 90% success rate at 1024x1024, even more when hires fix is used to reach 2048x2048, and even more if our included Ultra High Res Fix is used reaching 2576x2576. All the images shared have been made with it and SDXL Lightning checkpoints, without the assistance of inpainting, LORAs, adetailer or any other trick, just MasterPrompt and RealVisXL V4.0 (including the artistic images).

MasterPrompt is a prompting tool to be installed at Automatic1111 web-ui designed both for the beginner at the Stable Diffusion world and for the expert. It brings the power of an over a year researched prompting syntax, parameters and other automations, carefully developed, transformed into a program or algorythm capable of making high quality images with just a few clicks. This is the tool I have always been using to make my own images while I was developing it!!!

You can use it to directly generate whatever you want to see in a few seconds, but also as a steping stone to attempt further research as you can easily control any part of the prompt. You can freeze certain parts of it while leaving some others at random, retaining coherency, or even streamline from it to a production project. Watch it in action here (with install instructions included):

It will receive legacy support for SD 1.5 and SDXL based models, while it's current development will be focused on the SDXL Turbo and SDXL Lightning environment! Since January it is receivig constant updates, from weekly to monthly; from minor updates to increase the lists of avaiable combinations, to mid size changes as to support totally different setups such as artwork or even furry, to major further developments to support big different topics such as landscapes or warships; and everything in between!

Thank you everybody to help us come this far!

CHANGELOG:

> V-5.0.0 < Feral instinct!!! Animals and humanoid hybrids! Checkpoint selection automation and much more control!

Features:

- Added two new origins: "Creatures" that accounts for real and fantasy animals; "Feral" that creates human-animal hybrids and monsters with different succes by mixing Humanoids and Creatures prompts. This has caused a major change over the programs inner logic and provided chance for optimizations and simplifications.

- Updated to "weighted proportions" the way certain elements of the prompt get decided in the wizzard. It gives the user HUGE CONTROL over the amount of certain core elements of the prompt, for example allowing to change the probability of Humanoids or Creatures being prompted.

- Added the EXTREMELLY BUGGY "Checkpoint Automation" option, as it is powerful for producing image sets (but painful to use it for working and researching).

- Improved "Shot Distance" logic so that it is now more impactful over the final image.

- Great selection of initial creatures for the new category.

Changes:

- Hugely changed how many inner logic such as androginy work on the inside to accomodate the new "weighted proportions".

- Simplified and improved many lists.

- Changed "Shot Distance" to "Shot Composition" and removed "Medium".

- Changed and simplified logic and lists, such as "poses".

Bugfixes:

- Fixed "Shot Distance - Medium" being used for totally unrelated and obscure stuff.

> V-4.0.0 < Art & Thunder!!! Dramatic increase in quality, for SDXL Turbo & Lightning models! Added logic for Art and many new lists and functions!

Features:

- Added the "Type" selection and the logic to perform images either in Photography style or in Digital Art style!!

- Added the Hybrid Type which allows to blend art and photography images together!

- Added new "thunder" version to the program which has a HUGE increase of compatibility with MOST of the SDXL Turbo blends released and SDXL Lightning, but it's not compatible with base SDXL Turbo.

- Lightning and Turbo-Blends (Thunder) versions can produce 1024 images and reach 2048 using normal Hires Fix, while reaching 2.5k at maximun definition when using Ultra High Res Fix.

- Ultra High Res Fix works better than ever for other versions as well.

- Greatly improved automated generation parameters for ALL versions. Extra quality, yay!!

- Added readme with list of recommended models for the different versions and added it to the MasterPrompt's Manual. EXTREMELLY RECOMMENDED to use MP with those models when possible.

- Photographers/Artists are now considered style modifiers, as such, they now have a parameter for their chance of happening, similar to style LORAs, and their default chance of happening has been greatly reduced.

- Added a "verbs" logic file which allow to make different verbs happening across the prompt to be both version and logic dependent, for example to use "presents" for art and "records" for photography.

- Great simplification of the general prompt and logic, reducing the length by 13 tokens while increasing quality.

- HUGE improvement over the places prompts logic.

- HUGE expansion to places list.

- Great selection of initial artists for the new category.

- Great expansion and rework to clothes list.

- Great expansion and rework to nude bodies lists.

- Great expansion to poses list.

- Great expansion to qualities lists.

- Expansion of photographers lists.

- Simplified many lists.

Changes:

- Logic for files referencing the type of picture, the quality, or the author has been splitted to accomodate the difference between photo and art.

- Now adding the word "nonegative" to a negative list will prevent any negative from happening when it is selected (except those obeying other logics such as nudity prevention).

- Modified Poses logic and files to accomodate more options while getting rid of uneeded complexity: now front is everything that doesn't make sense from the back, there's no need for side or above, while fantasy category has been added.

- Modified Dressed logic so that it is more modular and similar to nude bodies logic. This adds the dressed-attractive, dressed-item and dressed-extra categories.

- Modified origin and photographers logic to avoid using some of them when prompting places, as well as editing improving the lists.

- Improved places lists.

- Removed the "Change Lists" readme as it was a pain to keep updated and didn't add enough value.

Bugfixes:

- Fixed an oversight that caused the sequence ', ;' to appear in the prompt under certain conditions.

- Added a logic to replace the word "selfie", if happened, for the word "photograph" in pictures from the back and from above. Expanded it for art as well.

- Fixed a bug that caused the wrong hybrid subject to be prompted ocasionally on the "places" prompts

> V-3.1.1 < Hotfix #2

Changes:

- Reverted having a resolution set in the automated parameters, as it resulted to be inconvenient and confusing.

> V-3.1.0 < Ultra High Resolution Fix!!! + Landscapes Update!: Ultra High Res Fix is back! And as part of Unprompted's program logic!

Thank you a lot ObezTheGod for your efforts in bringing it to life!

We also have the option now to make places as subjects, and the logic for defining lighting and type of place has greatly improved both as for subjects and as for backgrounds!!!

Features:

- Added Ultra High Res Fix!!! You can find everything about it in the System General Parameters.

- Added the "Origin" configuration. It allows to specify the type of subject that's gonna be prompted. At the moment "Places" and "Humanoid" are avaible, although "humanoid" stands only for humans right now. It is a core part of the image and will change a lot the behaviour of MasterPrompt depending on what origin it is trying to generate.

- Place Origin is fully implemented, allowing for LORAs and Embeddings the same way that humans admited them, including hybrid places too!!! The key distribution for places is "Interior" and "Exterior", as "Male" or "Female" was for humans.

- Added the new prompt category "Era" which goes along with the Places and modifies them.

- Totally reworked and expanded the lighting logic to make it an element of places. Lighting and time of the day as well as some special events that relate to illumination are mixed together in the category.

- Added "fantasy" to the "Reality Switch" at the Special Settings to force ONLY fantasy or unreal themed decission. VERY unstable; it's automatically disabled if "Disable unstable combinations" is selected.

- Totally reworked main's prompt structure for even higher fidelity and quality in all versions.

Changes:

- "ultra" option has been merged with the "real" option from the Special Settings.

- Deeply changed the backgrounds and lighting lists, with lots of additions and some deletions.

- Greatly improved humanoid prompt structure for even more obedience, specially in prompting nudes and poses.

- You can install Ultra High Res Fix only adding a file to your unprompted's installation folder, but that will only activate a suboptimal version of the feature. To get the best out of it you also need to replace a file of your A1111 installation, check install instructions for more.

- Modified automated parameters for all versions

- Modified general prompt "amidst" word for "within".

Bugfixes:

- Now it is not possible for any "body hair" to be prompted at "back angles".

- Now it is not possible for "abs" and other such features to be prompted at "back angles".

> V-3.0.1 < Hotfix #1

Bugfixes:

- Fixed an oversight with the race logic file.

> V-3.0.0 < Super Convenience Upgrade!: Ultra Determinism, User Interface, Embeddings and LORAS Huge improvements + Toggeables!

Features:

- There's been made HUGE improvements to the user interface of MasterPrompt. Fancier than ever!!! All the options are handled by easy to understand buttons.

- The version folders are detected automatically! Read next Feature, there's no need to write anything anywhere really now! (except for some very advanced cases).

- Now LORAs and Embeddings are detected automatically!!! That means that there's no need to maunally write any file within the program at all. All you need to do is drop your LORAs and embeddings into the appropiate folders within your installation. The folders are provided with the download, but if you need to learn more about it check the provided readmes. However if you want the same LORA or Embedding to be used across different versions of the program you'll need to duplicate the file. (Whenever you install new LORAs or embeddings you must Reload the UI through the settings button for them to be properly detected).

- The new version dependent "xnetworksoverride" allows you to define certain LORAs and embeddings that can have hardcoded strength and triggerwords if they are ever randomly selected. There's no need to write anything there unless you want to get rid of chance based values for those. It's version dependent because for example you may want different values for a LORA wether it's beeing used on a SDXL or a TurboSDXL based model, etc...

- There's been a huge rewrite of the inner logic. Now the tool is 100% deterministic, which means that now whatever choice you pick with the parameters and variables will only affect for that part of the prompt, but any subsequent prompt particles and images will follow the same logic deterministically. So for example changing subject from male to female will not start changing backgrounds and other parts of the rest of the images from a batch. Now the only thing that can alter the prompts for a given starting seed is changing the inner files to add or remove files. This is a very important change because it allows for bullk tests of any specific changes getting consistent output.

- Advanced Variables have gotten a "Toggle" checkbox. Ticking it removes that variable from happening on the prompt. It's the same as setting it to "". Bear in mind that not all variables can be toggled off in meaningful ways. This is an out of scope use, we have left acces to all the variables instead of only a subset of them for if you wish to rewrite the prompt files and use them differently.

- Added a button to the General Parameters to use system variables such as CFG from a version dependent file instead of using the A1111 interface. This allows both for randomization and also to quickly jump into generation as fast as clicking the Generate Shortcode button. This means that just by changing the version parameter, if this is allowed, all the needed params will change for you to work in a sdv1.5, SDXL or SDXL Turbo environment. Batch count will always need to be adjusted manually, as well as activating or deactivating Hires Fix.

- Added a whole bunch of new words for different list files to Turbo, SDXL and SDv15 versions.

- Added a small chance for SDXL Turbo prompts to have certain negatives.

- Added user-config file to the installation to prevent certain disregardable error messages from cluttering the console. Backup any changes made to that file before overwriting.

Changes:

- The huge rewrite of the programs logic has gotten rid of many of the variable clutter it had and uses more sophisticated ways to handle everything. This should result in it being more parsable as an human in order to edit it's logic files if desired.

- Because of the previous reasons, though, it is not backwards compatible, so any user made file edits will need to be done again.

- SDXL Turbo Upscale default factor has been reverted back to 1.334 from 1.5.

- Unnprompted's Error logging is turned off by default in the provided user-config file to prevent console clutter in cases where users haven't got at least 1 LORA and 1 embedding for every possible subfolder. Also it's configured to have the template open in load for ease of access.

- Manual Override is completly deprecated and not needed.

Bugfixes:

- All form of user dependent loops have been removed. This means there should not be any infinite loop problems anymore.

- Fixed some missguiding descriptions on the tooltips.

> V-2.1.0 < Resolutions improvements

Features:

- The user can now use different resolutions for different versions of the program. (All the features refer to using the "mixed" resolution General Parameter).

- Hi res fix upscale factor is also version dependent, and is prepared to be randomized if desired, although we only provide one base value, as we don't find a reason to use more than one. Keep in mind that high values can lead to out of memory problems.

- The Hi res fix upscale factor can also be toggled to obey the standard slider, overriding the above version dependent logic. Keep in mind that high values can lead to out of memory errors.

- The version dependent resolutions also allow for different values depending on wether Hires Fix is active or not, although we provide them the same on the base files.

Changes:

- Resolutions have been moved to the versions, this means that you can have different resolution choices depending on what version of MP you choose to use (new folder named "resolutions" contains the files).

- Added the "Readme-Known_bugs.txt" file to the readmes. Check it out to learn about some unexpected behaivours.

> V-2.0.0 < Resolutions, LORAs, Embeddings and Plug&Play Versions

Features:

- Now Stable Diffusion versions are handled as a dropdown list within the program. That way also final lists are chosen within the logic files and more easily managed in case that the user wants to modify them.

- Making custom supported specific models or workflows is easier than ever, as you just need to add an entry to the "Aine-Master_Prompt" function for this part of the code: "[set version new label="General Parameter" ui="dropdown" choices="turbo|sdxl|sd15|<YOUR_NEW_MODEL_GOES_HERE>"]turbo[/set]", copy and rename accordingly (to match the exact name of your new option) one of the prompts folders, then change any of the inner files to your liking. Thus, you can also share any of this versions by sharing that main folder and requiring to edit one single line of the code! (There's a readme explaining it better)

- Added basic Resolution randomization support both with and without Hires fix. Choosing "mixed" General Parameter will allow for 3 mixed resolutions to be generated. It seems that the first image from the batch uses some of the configured parameters instead of the logic, and that the resolutions for an image are decided on the previous image generation. Also this setting changes images that happen to have the same resolution even if they have the same prompts. More research and bugfix is needed, yet the feature is powerful enough to be used. I suspect it is related to how Unprompted or Stable Diffusion treat batch images.

- Added embeddings compatibility to the subject logic, there are some example embeddings in the list, but the user MUST donwload the files separately from the source, and should update the list to use any other personal embeddings he may want to use. You can choose to prompt female or male embeddings (including the faked mix type similar to the "fakefame"), or you can pick the "trueall" subject to have a little bit of everything. (There's a readme explaining it better)

- Added Style LORA handler. You can toggle on and off any LORA from happening during the image generation for style manipulation and you can decide the chance of this happening and the general strength of the used LORA. Keep in mind that you need to download the LORAs separately and that this is an experimental use of the tool, so output quality will depend a lot on the LORA (there's a readme explaining it better). As with the previous case, this setting changes slightly images that happen to have the same prompts even without any LORA prompted. More research and bugfix is needed; this is an experimental option, I suspect it is related to how Unprompted or Stable Diffusion treat batch images.

*While using high chances of LORAs happening you may want to rise the Setting: Extra Networks -> "Number of Lora networks to keep cached in memory" to something like 4-6 if your GPU can handle it, otherwise you'll have some deadtime between images as the new LORAs get loaded into memory.

- Improved the androgyny logic, now androginous subjects can be prompted without toggling the rest of the androgyny system, they are independent.

Changes:

- There's been a big simplification of the code using a relatively new feature from unprompted, this should help a bit with making user modifications of the program. All the logic files now just handle the final call in a 1 line of code.

- The logic for random colors (and materials) now allow for different colors being prompted on the same image and has the structure to allow for an unlimited amount of new colored/materialed elements; although no such a prompts are tailored yet. Also this makes that picking an option with colors doesn't change RNG. Many of the variables for the zcolorlogic are unused, just ready for future updates without changing RNG, if more are added they'd change RNG. Picking a color during variable selection would still fix all the colors to be the same one.

- General folder was removed in favor of dedicated folders for each element (colors and materials).

*Note that all these changes and features will require some tweaking to adjust any previously edited logic file, as this is not a backwards compatible update for logic files (word lists should be fine unless they used "calls" on their own).

Bugfixes:

- Fixed a crucial bug on the Determinism Fix I had implemented that caused all the images after the first one in a batch to use the same prompts, it didn't matter the seed. Hence it's strongly recommended to delete all the previous installation files and make a fresh install of this update.

- Fixed a typo that caused wings to be prompted from back angles, which they shouldn't (SD models don't know how to make them propertly in unbounded scenarios).

- Removed minor leftover code from previous versions.

- Fixed an oversight that rarely allowed for a few poses to happen under unintended circunstamces.

- Fixed an oversight that allowed for androginous subjects to be prompted in circumstances where they shouldn't.

> V-1.0.0 < Release!!!

描述:

This is prepared to work with any model based on SD1.5, SDXL or with SDXL Turbo. In the images you have examples for using the 1.0 version. Ignore the downloads from this version itself, they are deprecated example wildcards.

训练词语: