Changelog for 0.7.3 beta
- Stable Diffusion prompt builder now supports weights. I suggest github.com/hlky/stable-diffusion benefiting from Gradio GUI. Run that and enter the prompt (including the weights) copied from promptoMANIA to the input field on top.
- Menu items, including the prompt builder local menu and the main nav will now close the menu when clicked/tapped.
Changelog for 0.7.2 beta
- Added the third set of example images (landscapes) to Midjourney prompt builder. They now have 3 different example images (beautiful symmetrical face, sphere, landscape).
- Added the third set of example images (spheres) to Stable Diffusion based models. They now have 3 different example images (beautiful face, sphere, landscape).
- promptoMANIA settings became Base image and it is on top, instead of bottom, so they are even harder to miss and easier to understand.
- The beta model was removed from Midjourney, therefor I have removed the
--betaparameter added in 0.7.1.
- Created a more robust error handling solution with some graceful fallbacks for the user. If you still experience a crash or server error, please report it with a screenshot. Thank you.
Changelog for 0.7.1 beta
With this patch, you can select the beta model in Midjourney, and if you do, a new set of example images and parameters will be available!
- Now with full support of the beta model in Midjourney. Setting the
--betaparameter will disable and hide most other parameters. This is intentional, as those parameters can cause errors with beta.
- Added the first two set of example images (landscapes and beautiful faces) to Stable Diffusion based models, including the new beta model of Midjourney. I'm thankful to Guillaume Audet Beaupré for this, and all his help with SD.
- promptoMANIA settings are now open by default, so they are harder to miss (cog icon is also rotating while they are open, and stops when closed)
- Many speed improvements to make sure promptoMANIA will stay fast with the added complexity
Changelog for 0.7.0 beta
promptoMANIA reaches beta stage!
- The most significant new feature in beta is that we can choose between different example images for the prompt parts, thanks to our new research assistant, Tuleyb Simsek. You are awesome! This also means I am no longer using any part of the Willwulfken Midjourney Styles and Keywords Reference. It was helpful in the alpha, it saved time back then, and for that, I'm grateful, but it is now time to move forward.
- Aside the sphere we have another example, the beautiful symmetrical face, and we have the ability to add more. (Landscape coming soon.) You can select the example image from promptoMANIA settings (bottom of the prompt builder)
- Four-grid example cards were replaced with single image cards, this will be user-tested extensively in the next few weekes, hoping to improve understandability and the whole information architecture.
- Generic prompt builder added to the prompt builder selection. You can use this for DALL-E 2, Disco Diffusion, Craiyon, or any other diffusion model.
- DreamStudio prompt builder added to the prompt builder selection, supporting the beta.
- The prompt builder bar (on top of the page) has a new engine. It has a neat local menu as well, jumping to parts of the prompt builder and opening/closing them with a click. It can be found in the beginning of the bar (top left corner), with the icon:
- Beta upscaler
--upbetais now supported by the Midjourney prompt builder. It is a parameter, just below the Light upscaler. I don't want to make it more emphasised until it's no longer beta.
- Made the prompt builder bar (on top of the page) have editable input. This does not have two way data binding, so edits made there will not be reflected on the rest of the page, it is more for "finishing" touches.
- Fixed a bug where it was possible to give individual weights to Midjourney image inputs. That is not supported by MJ yet, so it was removed.
Changelog for 0.5.2
- Grid splitter has been released, so you can separate grids of images generated by Midjourney or Stable Diffusion into individually savable images.
- This patch addresses the parameter changes in Stable Diffusion:
-ghas a maximum of 4, instead of 9,
-shas a maximum of 50, instead of 150.
- Every prompt part and every parameter now has a default, meaning that the most likely value or the default from the given diffusion model. For weights this is 1, except image weight in MJ.
- Changelog got a permalink: /changelog/
Changelog for 0.5.1
- This is a hotfix less than 24 hours after 0.5.0 to address the first set of bugs reported.
- SD parameter for custom weight and height had a bug. This was reported by Supa Mike, thank you so much.
- SD parameter for CFG (classifier-free guidance) scale factor has been updated to allow negative values. Please be aware that negative values mean the opposite of the prompt.
- Image weight became the first in the Midjourney parameter list, because people had hard time finding it, and it seems to be popular with a small portion of the user base.
- Gideon Reiss reported that the button for reference image prompt parts are still hard to find. This has been improved. Hopefully everyone can find it easily after this patch. Thanks Gideon.
- The name of Hayao Miyazaki (宮崎 駿) was spelled incorrectly, this has been fixed. Thank you Devin Zacharias for reporting it.
- Adding a value to a numeric input inside a checkbox (like seed) sometimes did not set it to checked, so users had to manually check the checkbox at the begining of the row. This has been fixed.
Changelog for 0.5.0
- Alpha was extended because the prompt builder engine was rebuilt from sratch. It has tons of new features, and even more possibilities for the future. The same modules can support any number of Diffusion Models. This was needed because starting with this version there will be full support for Stable Diffusion (SD) by Stability.ai.
- if you are using Midjourney, you will also find a much better prompt building interface, which is more extensible, and it will be easier to accommodate future changes by the Midjourney team.
- Animations are added to enhance user experience (cards flying up to their new place or going back to details/styles when deleted)
- Image prompts are reworked, you still need to find external host for the image, but the UI is a bit nicer and harder to miss.
- In Midjourney prompt builder weight zero means that the prompt part is discarded. Now the prompt builder reflects this. (It will not add the prompt at all, instead of just adding ::0)
- New prompt-builder page, starting with the selection of the diffusion model. You can bookmark the prompt builder for your favourite model, no need to go to the selection page every time.
- Intentional camera movement added to Details > Camera > Scenes. Huge thank you to one of our Patrons, John Stevenson for the suggestion.
- Feature roadmap removed. The diffusion model scene changes, and certain things jump on top of the list anyhow, like supporting SD (Stable Diffusion).
- All prompt parts now fully support the Grammarly spell checker. Use it if you need spell checking, as it is pointless to try to build a better one.
- Text detail card removed from Midjourney due to poor results. Hopefully there will be a way to consistently put readable text on Midjourney images, but for now this is highly unstable.
- A new icon was created to show a hanging weight with two colons
(weights in Midjourney are added with
- Svelte updated and all dependencies checked. Playwright thinks that everything works as intended, but if you find any issues, please report them!
Changelog for 0.4.7
- We have a new sponsor, Vitropep the startup that will make syringe injections disappear!
- Reworked parameter area, got a slightly improved design, image weight was also moved there because it is a Midjourney specific parameter after all
- Added the
--chaosparameter to Midjourney prompt builder
- small bugfixes
Changelog for 0.4.6
- Another small, but significant change: Added lookback to prompt builder engine,
so now if you add a weight, the previous prompt will get
::. Most users, according to my data will prompt
forest, green::-0.5so something similar, if they want a forest that's not green. Midjourney will understand this as an image that contains no forest and it's not green. So promptoMANIA will now fix this to be
- Minor scss and ts improvenets, and small bugfixes
Changelog for 0.4.5
- A hotfix to replace
,when no weight is used, as requested by the Midjourney team.
- The weight engine supported this change, so it is a relatively small change, but can cause bugs, so please test it, and get back to me with feedback.
Changelog for 0.4.4
This release has been adopted by Master Jinlin (金玲). Thank you ever so much!
- Added the first Midjourney interactive tutorial and built the first version of the interactive tutorial engine
- Added text on image prompt using starshadowx2's method with a fix. You can find it in the details as the last item (currently). A small fix (compared to raw starshadowx2) was needed, becasue high stylize and low version numbers have a tendency to ruin this. So when you set it first, the stylize will be set the minimum and version to maximum. Please give me feedback if this works for you, or even better if you have a more advanced text to image solution.
- Awesome things in the code became even more awesome - I love refactoring and speed optimizing in the sunshine!
Changelog for 0.4.3
- Midjourney parameter selection was redone; version number became a slider
- I made sure that quality and quantity sliders are not shown if engine version 1 or 2 is used, to avoid confusion, as those parameters don't work with the older engine
- The sliders now work similarly with the rest of the parameters, with checkbox to enable/disable them. The module was created and cowboy code removed.
Changelog for 0.4.2
- 25th of July Midjourney update fully supported!
--ssliders and numeric inputs added
- Made sure that Quality slider can only have values from the array of
[0.25, 0.5, 1, 2, 5]
- Old Midjourney algoritm parameters added
- Deprecated parameters removed:
Changelog for 0.4.1
- hotfix to prevent a rare but annoying crash in 0.4.0 (released a few minutes ago)
- made the separator used in Midjourney
::a variable instead of a constant, this is the first step to reintroduce support for other VQGAN and CLIP based AI art builders, such as DALL-E 2, Disco Diffusion, Hypertron, Crayon, etc.
- previously users could go beyond 100% or below 10%
--stopvalue, now this has been fixed
- small bugfixes and typos here and there
Changelog for 0.4.0
- This is the last alpha version introducing new features. After this I will focus on getting ready for the beta version 0.5.0. The plan is that the beta will be released in August, and should be relatively bug free. If you find any bug or typo, please report them.
- Weight component done! Now pretty much anything can have weight (starters,details, styles, images, etc)
- Based on real user testing and tremendous patron feedback, I changed the 'detail drawers' introduced in 0.3.7 into a new solution. This will be tested in this release, and if it works, I will keep it at least until the betas.
- Added new styles and details: Weta Digital, H.R. Giger, Frank Frazetta, David Mann, Unreal Engine, Pre-Raphaelite Brotherhood just to name a few. Can you find the easter egg?
- Ability to use three subjects added, but it is not recommended with Midjourney at the moment. If you make that work, please share the prompt so we can learn from your mastery.
- Added the Midjourney parameter
--stopwith slider and numeric input to stop the rendering early. With this, all currently known and supported parameters are handled. Remember
--stylesupport was dropped because it was never officially supported, causes issues, overpowers prompts, and is deprecated.
- Refactored the prompt builder to improving the typescript for the betas
Changelog for 0.3.7
- The new engine is now filled with content, just the details would be over 16000 lines if the .json would be expanded. This added a substantial increase in possibilites and use-cases (but also server load, so please consider supporting this project)
- Added detail drawers, so you only have to open detail types you need
- Added images from Midjourney Styles and Keywords Reference by willwulfken as a temporary solution until I create the final images for the project. Thank you so much for making a public repository and allowing me to merge it into promptoMANIA. You are awesome!
- Added many painters, huge thanks goes to Fabio Gamberini for his heroic work so far as an art consultant
- lodash deep clone implemented to make deep cloning of JSONs more efficient, now we have many json files some of them over 16k lines, so this was a must before the new features could be released
- Image weight bug reported by Stuart Wilson is fixed, thanks for the report
- Fixed the style bug reported Richard Barker, thank you for the report
- From now on
--stylesupport has been dropped, as that is officially unsupported by Midjourney. It was introduced in 0.3.0 but seemingly it was a mistake.
Changelog for 0.3.2
- Images are now in Amazon S3 buckets and delivered to you through Cloudfront CDN (meaning everything will run faster)
- did some ui improvements (example images became smaller) as a first step to accomodate waaaaay more prompt building blocks with images and weights
- added baked-in weight support to be able to add back the most in-demand detail of
watermark::-0.3 blurry::-0.3 cropped::-0.3, before the weight engine is fully implemented
- custom style and custom detail became text area, the questionmark placeholder was retired for now
Changelog for 0.3.1
- Image prompts are now supported!
- Including the experimental image weight
- Added new artists
- Fixed the image cards on smartphone, now they only show two images instead of the full 4, but hopefully this will make the experience better
- Fixed typoes and a few minor bugs, including a bug making impossible to have an image prompt without a text Prompt. Now you can have an image prompt only prompt built.
- a bit of frontend speed improvement (though not noticeable unless you run this on a toaster)
Changelog for 0.3.0
- New prompt builder engine: flexible, extensible and feature rich
- Example sphere images added to provide visual feedback for style and features
--testbecame default option, beacuse faces, scenes, and creatures are less distorted. Images are more colorful, more “pretty”, and “cleaner looking”
- Fixed some other minor bugs and typos
Changelog for 0.2.9
- Added "read more" functionality to the site, the first use-case is this changelog.
- Getting ready for 0.3.0
- Fixed a bug with popper not working properly for certain edge cases, ended up rewriting a bit of Popover.svelte functionality
- Fixed some other minor bugs and typos
Changelog for 0.2.8
- Redesigned the prompt building engine to support most parameters to
- Added preset ratios with a bit of background information about them.
- Copy buttons now copy the whole command to the clipboard. This includes
/imagine prompt:command at the beginning and all parameters.
- Sizes became a section in the prompt builder, with many preset aspect ratios known to work well with Midjourney
- Now you can specify the random seed to set the pseudo-random noise used during the
initialization of the generation. Please be warned that there is no enforced natural number
validation at the moment, just a
min="1"attribute. Negative random seed produces an error with MidJourney and zero means "none" for seed.
- The dropdown menu is populated with sections for quick navigation within the prompt builder
- Separated this change log into a new library to enable future improvements
- A nasty copy bug squashed - in obscure browsers, like the Facebook Apps built-in browser copy didn't work, now it does
- Fixed a bug where
· and/or ·was not showing on certain devices. Now it almost always does, and worst case scenario a
·will be shown to make sure people get that they don't need to do BOTH inputs.
Changelog for 0.2.0
- The design was changed to a dark theme, custom-built by me on top of Bootstrap. According to most patrons, this is the preferred color scheme and saves energy. Your eyes will also be less strained.
- Shifted focus to Midjourney, instead of trying to support many VQGAN and CLIP based AI art creators. The prompt builder should still work with DALL-E 2 and most VQGAN and CLIP based implementations, like Hypertron (v2), Disco Diffusion (v5.4). Midjourney focus means that specific features will be developed and tested for Midjourney. However, suppose the community needs focused support for DALL-E 2 or any other AI. In that case, I could create a target choice selector for the beta, as the builder is designed for flexibility and extensibility.
- Added a mini-tutorial to use the prompt builder at the end of the form, more complex tutorials will be published later
- Fixed a bug with the early implementation of text weights. In the future Midjourney style
weights will be used:
::weightValue- thank you Luca De Marini and dezm0n for reporting this bug
- Added preset flags and also a checkbox to use custom as suggested by Adam Pizurny and Valics Lehel.
- Symbolic style added - Jack from Discord suggested it
Changelog for 0.1.8
- Fixed a nasty deployment bug generating 404 error on existing pages
- Added a better 404 page with a suggestion on how to avoid, if similar issues appear in the future (hopefully they won't)
- Valics Lehel reported that there was no separator between the two starting prompts if both
are used (A and B). Now there is a comma [
- Implemented page speed improvements, now the prompt builder has a PageSpeed Insights score of 95 on mobile and 97 on desktop
- Added this changelog
Changelog for 0.1.7
- First public alpha release