Welcome to AI Image Central's new forum. You can register for the forum with Google or you can verify your email the old fashioned way.
*** If you don't log in, you can't see any of the attached images in the messages. ***
New updates:
- Now using SDXL Lightning with a choice of 19 models! 10-45 seconds including model load time!
- Images are now sorted from newest to oldest for easier viewing.
- Rewrote the code to use thumbnails for preview images instead of loading the entire image. Massive speed increase!
- Added this new forum
This site has been designed with desktop users in mind as it's a graphics site. If you are viewing from a mobile phone
then please set your browser on the phone to display in desktop mode. The site will look proper with that setting turned on.
New Update:
- Added IP Address to log file along with prompt information in case I need to ban someone who is not following the rules.
- Keep in mind this is a public site and only tasteful nudity is allowed.
New update:
After spending hours trying to get it to work correctly, I was finally able to include a remove button in the Requested area.
So if you don't like whatever you created or you accidentally created an image that goes against the rules and need to get
rid of it, now you can. I don't actually delete the files but move them to another folder for review in case you accidentally
remove something you didn't mean to. You can leave a message to see if I can recover it. I will be emptying the folder
often so if needed make the request ASAP or you may be out of luck.
New update:
Due to people creating illegal images I have had to find ways of limiting access. I have added code to now prevent people on
VPN's from generating images. There is no message, the buttons will just not work. You can still browse and download images,
but creating images will not be available. So if you want to create your own images, you cannot be on a VPN. If I still have
issues with people not following the rules and doing illegal activities then I will be adding a login page that will require a sign in
with Google in order to access the site. I already have the code ready if needed.
I have been playing around with SDXL and Flux along with Ultimate SD Upscale... The upscale is simply amazing... But with Flux it's spectacular...
Not 100% still, but it's a lot closer then SDXL. But it's also much slower. It takes 630 seconds to upscale using Ultimate SD upscaler with Flux.
Check out this bedroom... Can you even tell it's computer generated.. Imagine if it was the size of a photograph.. It would be so realistic.
I also tried the 4bit version of the Flux dev model but it's just as slow as the 21gb modified fp8 version I'm using so I see no benefit to using it.
I think the images were subpar also for Flux. I was hoping to find a way of making Flux fast enough to include it as an option but that's just
not possible on my hardware it seems. I made an API and everything to run flux but somehow it seg faults (clean exit with no error message!
but need to restart) and it's slow even with my 6 step "lightning" modded version. What's weird is I can run it just fine from my workflow
which is exactly the same as the API version. Some bug in ComfyUI API interface it seems...
New update:
I have added a new search engine that allows you to do keyword searchs on image descriptions. Still trying to get the search the way I want.
For now your limited to entering single word searches unless your searching for two or three words in the same order.. like 'White Hair'
or 'black and white' will match so long as the words are beside each other in the right order. The results cover all folders except Requested.
New update:
Search engine has been finished. You can now search with multiple words and select if results match any word or all words.
New update:
I have done a bunch more work on fixing up both the Prompt search and Image seach functions.
You can now select three different search types making it easier to find what your looking for in both.
New update:
I added a group chat feature to the main index page. Click the little message bubble in the bottom right corner to bring up the chatbox.
If someone sends a message, you will hear a ding and see the chat icon shake for 1 second.
If 'Do Not Disturb' is selected (Turns red when selected), the ding will be muted.
New update:
I have been working on improving the chat feature...
Now you can drag and drop jpg,png,gif, and webp files into the chat (even from the website) as well as text.
You can also right click the images to display in new window. After a few days of hard work the chat now stays
populated and remembers previous settings across folders and refreshes! Chat and generate or browse at the
same time!
Found flaws with the image drag and drop so might not work for you. You may just get a broken link image.
Working on getting it working 100%
Probably adding more customized buttons... Verifying the DND works... (Works for me)
And still trying to figure out how to make the searches work 100% when you hit search.
Stupid thing sometimes brings up previous results instead of doing the search...
New update:
I have done yet more changes... Changed the look of the website to make it a bit more modern.
Got rid of the secondary menu so it's more stream lined and easier to get around.
Locked the title and menu so they no longer scroll.
Created all new buttons and folder system.
New update:
The group chat is almost perfected. You can now send MP3 files as well as PNG/JPG/GIF/WEBP.
I believe the only bug left is getting the server/client to work properly as far as retrieving the stored files after the
browser is refreshed or the webpage is changed. Images work, but MP3's even though they are being saved in the
folder are not being recognized and used after the refresh wipes the data. I have routines in place that save the
settings and reopen with your previous settings so you can still chat while browsing the website.
New update:
I was finally able to get around the API issues and get Flux working. So now you can also run the Flux model which is
good at adding text. So if your trying to design something with text in it, that is your best bet. It takes around 50 seconds
per image because I limited the size to 960x544. Flux is time intensive even using 'lightning' type version. Just to go to
1280x720 takes 125 seconds, about 2 1/2 times what it takes for 960x544! You can use the upscaleX2 feature in the
Requested folder to scale to 1920x1088 if desired.
One of the users here created this video from art they generated here. Pretty amazing stuff!
https://www.youtube.com/watch?v=Z1bt_-CHDb8
New update:
I have removed the audio messages from the group chat. You can still drag and drop .mp3's to add to the music player but
they won't show in the chat (Only in the MP3 Player). The code was downloading all the songs automatically using a bunch of
bandwidth for no reason. So now they will only download if you open up the MP3 Player because you want to listen to music.
The drag'n'dropped MP3 will populate to the end of the list of MP3's in the MP3 Player. Note you cannot drag'n'drop into the
MP3 Player, only the chat box. Right now I have limited the music to around 25 songs but I am working on other ways to
incorporate a much bigger selection. I manually change the selection of songs every few days right now.
New update:
I added the Spanish language to the site so you can now switch to Spanish if you like.
I may add other languages but it's a lot of work... And I'm not sure what languages the models understand so will investigate.
New update:
I have added a translation to English that is automatically run now if you type in a prompt on any of the Spanish webpages.
So you can enter the prompts in your native tongue, and it will still work correctly.
The model works best with English and without the translation it won't match the same outputs as in English.
I will have to see how well it works before I decide if it's the best way to move forward but I believe so as this way I
can add whatever language and it does not matter because the prompts are converted to English for the models.
New update:
I have now also added Hindi as a language option. Again, like Spanish, the prompt will be converted to English before
being put through the model for the best possible results.
New update:
I have now also added Traditional Chinese (ZH). Like the other two languages, the prompts are converted to English in order to get the best possible results.
So the website now supports English, Spanish, Hindi, and Chinese. About 3.5 billion people?
I have updated the rules, so please read them or read them again.
I have also provided some information about some different features on the site like the music player and group chat.
You can view the Rules here: https://aiimagecentral.com/forum/index.php?topic=4.0
New update:
- Added SD 3.5 Large Turbo model.
- Added Rules Tab to main screen.
- Added SD 3.5 Medium model.
- Added multiple step options for both Flux and SD 3.5 Large Turbo (4,7,10 steps)
New update:
- Changed the dropdown menu for SDXL to make it look nicer.
New update:
- Fixed the Spanish and Hindi languages to now use the new layout. (Still 3 more languages to go...)
- Added Stable Diffusion 3.5 options to Spanish and Hindi languages.
- Created Rules for Spanish and Hindi languages.
New update:
- Fixed the Traditional Chinese and French languages to now use the new layout. (1 more language to go...)
- Added Stable Diffusion 3.5 options to the Traditional Chinese and French languages.
- Created Rules for the Traditional Chinese and French languages.
New update:
- Fixed the Arabic language to now use the new layout.
- Added Stable Diffusion 3.5 options to the Arabic language.
- Created Rules for the Arabic language.
New update:
- Added Image search to all other languages in that language. Translates whatever language to English and then does the search.
New update:
- I see a lot of issues where people just simply don't READ. I posted all over the website that you can't do text in an image with SDXL but people keep trying.
So I have changed the default model to Stable Diffusion 3.5 which is newer and CAN do text. Using more steps will help with getting the text correct.
Text is still limited.
Stability AI has released a prompting guide for SD 3.5. You can find it here: https://stability.ai/learning-hub/stable-diffusion-3-5-prompt-guide
I have also included another prompting guide previously in the forum which you can read here: https://stable-diffusion-art.com/prompt-guide/
You can find a guide to Flux prompting here: https://fluxaistudio.com/blog/prompt-guide
New update:
- Added Another set of options under SD3.5 for the Large Turbo model to also generate from the 8GB GPU.
So you can run the Large Turbo model from both the 24GB and 8GB GPU's. The 24GB GPU uses the fp16 T5XXL
text encoder and the 8GB GPU uses the fp8 T5XXL text encoder.
I also changed the sampler for the SD3.5 models to Euler as it appears to give better results. Thinking about possibly
adding more options like selecting the sampler and seed instead of being hardcoded. Maybe add other schedulers...
Like Beta. Any thoughts?
New update:
- Add sampler and scheduler options to the Stable Diffusion 3.5 models.
New update:
- Added sampler and scheduler options to Flux models.
Here is a little guide for Flux model selection I found on Reddit:
These are the only scheduler/sampler combinations worth the time with Flux-dev-fp8.
🟢=Good 🟡= Almost good 🔴= Really bad!
Here I have compared all sampler/scheduler combinations by speed for flux-dev-fp8 and it's apparent that scheduler doesn't
change this much, but the sampler does. The fastest ones are DPM++ 2M and Euler and the slowest one is HeunPP2.
From the following analysis it's clear that the scheduler Beta consistently delivers the best images of the samplers.
The runner-up will be the Normal scheduler!
SGM Uniform: This sampler consistently produced clear, well-lit images with balanced sharpness. However, the overall mood and
cinematic quality were often lacking compared to other samplers. It’s great for crispness and technical accuracy but doesn't add
much dramatic flair.
Simple: The Simple sampler performed adequately but didn't excel in either sharpness or atmosphere. The images had good
balance, but the results were often less vibrant or dynamic. It’s a solid, consistent performer without any extremes in quality or
mood.
Normal: The Normal sampler frequently produced vibrant, sharp images with good lighting and atmosphere. It was one of the
stronger performers, especially in creating dynamic lighting, particularly in portraits and scenes involving cars.
It’s a solid choice for a balance of mood and clarity.
DDIM: DDIM was strong in atmospheric and cinematic results, but it often came at the cost of sharpness. The mood it created,
especially in scenes with fog or dramatic lighting, was a strong point. However, if you prioritize sharpness and fine detail, DDIM
occasionally fell short.
Beta: Beta consistently delivered the best overall results. The lighting was dynamic, the mood was cinematic, and the details
remained sharp. Whether it was the portrait, the orange, the fisherman, or the SUV scenes, Beta created images that were both
technically strong and atmospherically rich. It’s clearly the top performer across the board.
When it comes to which sampler is the best it's not as easy. Mostly because it's in the eye of the beholder. I believe this should be
guidance enough to know what to try.
Found a chart for SD 3.5 that shows good combinations for Stable Diffusion 3.5 samplers and schedulers:
New update:
- Added sampler and scheduler options to SDXL models.
Also take note, the samplers and schedulers I believe work best are near the top of the list but I have included many that mostly work for the different models.
A couple combinations will give bad results. Check the charts in the previous messages.
New update:
I was never really happy with the turbo Flux quality; it takes forever and the quality for turbo is quite frankly not anywhere near as good as the pro version.
However, I now found some modified models that give really amazing results very similar to the pro model. It takes much longer
than any other model taking over 2 minutes to create a single image, but the results are excellent.
So you now have the option to create flux.pro quality images if you don't mind waiting. I also increased the allowed prompt length
for Flux to 4000 characters as it does understand much longer prompts.
The options are for 20 and 25 steps. I also included a version for text in the image and a version without text in the image.
If you're not trying to add text to the image, use the [NO TEXT] options as the model uses the T5XXL 8fp leaving more room for a bigger model.
If you're trying to add text to the image, then use the [TEXT] options as they have the better T5XXL 16fp text encoder.
Some examples at 20 steps:
New update:
- Added a Stable Diffusion 3.5 Large model with the same resolutions as Flux (960x544/1920x1088)
- Added the options for 20 or 25 steps.
- 20 steps take 90-120 seconds depending on if the model is already loaded or not.
Some examples:
I am seeing people typing in other languages on the main page in English, so I decided to auto detect the language on the English pages
as well so that you can basically type the prompt in any language and it will be translated to English. I was having issues with the
charmap because of people typing in weird characters that are not part of the English language causing the prompt to fail. Doing
this also solves that issue. It really is amazing how difficult they make it to fix issues with Unicode characters!
New update:
I added progress indicators now for both GPU's so you can see the progress or see if it's busy. I added to all models and all languages! (18 PHP files!)
I fixed the generated image output so it's much faster to update the image. Before it was hit or miss as far as how long it took.
And I added a colorful background behind the title area.
New update:
I decided to change the menu to make it look newer. Finished updating all languages.
Let me know what you think.
New update:
- Made nicer buttons for Previous/Next.
- Made nicer buttons for 'Face Fix', 'Upscale X2', and 'Remove BG'.
- Moved 'Face Fix', 'Upscale X2', and 'Remove BG' to the full screen image view. (Click on the thumbnail in Requested)
- Fixed the Info window so the text wraps and made the window bigger to display 3000+ character prompts.
- Rounded the highlights and buttons to get rid of the ugly square look.
New update:
- Finally fixed an issue with CLS (Cumulative Layer Shift) at like .32 due to the slideshow. Found a new way to do it and now CLS is an awesome .015!
Supposed to provide a good user experience. Anything over .1 is marked as needs improvement. Anything over .25 is considered
an error. Hopefully now the site gets ranked better.
-Moved the sampler to the other side with the scheduler so the model is listed in the first dropdown menu now.
- Modified the page layout to display the page better and fit more in a smaller space like a tablet with a lower resolution.
- Changed the textbox's to better text input methods.
- Modified top right model links to now apply the full path from where you're at when switching models (Click the title image to go to the main page).
Screen capture of the test results:
New update
- Fixed placement of <title> <body> and other <meta> tags to enhance search results.
- Added latest changes to all English, Spanish, and French files. Still 3 languages to go.
- Changed the titles and descriptions to include either the model if on the main page or category and page number.
New update:
- Added menu bar with full menu to prompt search, image search, about, and rules.
- Modified the searches so that the results appear on the same page. Looks much better and you can see previous results.
- Added the ability to download the original PNG file by clicking the thumbnail preview on the generation page.
New update:
- Added the new ShuttleAI 3.1 Flux model to the Flux tab. Option for 4, 6, or 8 steps which is all it needs. Pretty fast compared to the other Flux models.
Once loaded, it's only 35, 45, or 55 seconds for 4, 6, or 8 steps. I can only load the fp8 model as the fp16 model is 23.8gb leaving no room for the
encoders. Results are impressive for the limited number of steps.
New update:
- Finally finished updating all the other languages to the latest format.
- Changed back to textbox from the new method I was trying. These boxes work better and can be expanded if desired.
If you use the back button your prompt is also still in the box unlike the other method.
New update:
I have created a new feature but have not decided where to put it yet. The new feature is Flux Fill which is a really
good inpainting model. It takes like 4 minutes to run and 24gb VRAM but the results are very good.
The purpose of such a model is to modify an existing image where you like the image except for some flaw.
You can use the Flux Fill model to mask the part of the image you don't like and then create a prompt to modify that area.
For example, you have an image of someone wearing a shirt you really hate. You can simply use the mouse to draw
over the shirt area, then type in the prompt 'white tank top' or whatever and then wait about 4 minutes for the results
in the Requested category.
The current link is: https://aiimagecentral.com/flux-fill.html
Here is an example where I changed the eye color to red..
Another example...
Added the button for Flux Fill. Flux Fill is very good at inpainting.
Added a NSFW detection model to block any inappropriate content.
Changed the NSFW detection model to a different one. The other one is sometimes blocking stuff that should not be.
This one I have set quite high but believe it should work better and it's smaller, faster, and should be more accurate.
The NSFW model is blocking more than I would like but so far, it's been the best option to block unwanted images.
A lot of times it seems to block an image if the body is somehow messed up so not a big loss in those cases since
the image was not very good. I will monitor the NSFW folder and re-add any images that look decent and should
not have been blocked. This mainly has to do with images of bikini's being blocked. I would prefer a model that
just blocks nudity so I will keep looking but the VRAM requirements need to be small, so the NSFW model fits in
memory with the loaded diffusion model to prevent reloading after every image which would slow down image
generation.
Added more images to the categories folders. Now up to 9,401 images with descriptions on the website.
Almost hit that first 10K! I still have 4,000 more images to go through and see which are good enough to add to the site.
Should be over 10K in the next couple days when I feel like going through more images... Already went through 6,000...
New update:
Since I now have a pretty good filter in place to stop the unwanted images, I have decided to try removing the VPN block.
Will see how it goes. I also removed the bans. Follow the rules or be right back to being banned.
New update:
- I was finally able to fix the font size on the textarea boxes for the mobile display.
- Changed the Group Chat box size to fit better on mobile displays.
- Added some ads to the website to generate a little revenue. May upgrade equipment eventually. But ad revenue is very low.
- Moved the chat icon and Group Chat to the top of the page on the right.
- Applied changes across all 6 languages.
I started a GoFundMe fundraiser to help raise money to upgrade the equipment on the website.
The GPU's are very outdated and the generation time can be greatly sped up with newer hardware.
If you would like to donate towards this goal, then please go here: https://gofund.me/9f6915af
I have updated the Rules/Help to include permission for free use of any images for company logo's or any work-related
purpose where there is no direct profit from the image or images. So you don't have to worry about using any images
for those purposes. For example, you can create an image for a presentation but not create an ad or a product you sell
with the images. Like mug graphics or t-shirt designs.
I decided to create a category for NSFW. If the image generated is detected as NSFW, it will be moved to the NSFW folder instead
of the requested folder and a message will display telling the person to check that folder and not show the image to everyone.
This way people can generate nudity, but other people don't have to view it if they don't want to.
I also combined the cars, trucks, and trains to a vehicles category to make room for the NSFW category and allow the categories
to fit better on mobile devices.
I decided since I modified the site to allow NSFW to add a couple Pony SDXL models. I removed the Crystal Clear model and added the
Lustify and Damn Pony models. These models are trained on generating nudity. Remember to remove any nudity or NSFW
prompts from the Requested category. Only allowed in the NSFW category!
I have re-instated the ban on VPN's and IP's that don't resolve due to a Pedo that won't take a hint and keeps generating child porn.
I have already submitted a report to the Cyber Crimes division logging all his activity and IP's for them to track him down.
But who knows how long that will take dealing with other police forces outside the country.
Most IP's resolve except for VPN's who try to hide so you should be able to still generate unless you are on a VPN.
If so, disconnect from the VPN in order to generate images otherwise pressing Submit will do nothing.
New update:
- Modified indexes, search pages, about, and rules to display correctly on mobile devices without desktop mode turned on.
- Applied above modifications to other 5 languages.
New update:
- Converted all the thumbnail images to the WEBP format from JPG as the file size is half the JPG file size allowing for even faster load times.
- Converted all languages to use the WEBP files and removed the JPG thumbnails.
New update:
- Added Age verification pop-up to the NSFW section. Once Yes is selected, the pop-up will no longer appear for that browser
unless you clear the cookies. If you access from another browser, you will be asked again as it's stored in the browser.
New update:
- Added some basic LORA support for SDXL models via a drop down menu.
Right now the selection is limited as I'm just making sure it works and I still have to change the other languages.
The trigger word is automatically added to the end of the prompt.
Current selection includes:
- Glass sculptures
- Stained glass
- Blacklight
- Renaissance
- Fine nude
- Greg Rutkowski style
- Wowifier
If you have any requests for any specific SDXL LORA's then reply here with your requests.
Added a bunch more SDXL LORA's... Now up to 37!
New update:
- Added a new model called Lumina 2. The model takes 2 minutes to generate an image. Does not do nudity or famous people.
- Added Babes by Stable Yogi Pony SDXL Lightning model. This is the third SDXL model added that is trained for nudity.
In order for the LORA's to work better and possibly provide more detail to the images, I have increased all the SDXL models to 12 steps.
If the model is already loaded, it still only takes like 10 seconds so no big increase in time but should give better results.
The site is now back up and everything is back to normal.
Added a new 2TB SSD and switched ISP's.
I added another model to the Stable Diffusion 3.5 section. The new model is from TensorArt and is called TurboX.
The model is 8 steps and takes about 72 seconds when not already loaded and around 40 seconds when already loaded.
The results look promising.
I have removed the SD3.5 Medium model as the quality is quite frankly the worst of any model I have ever seen.
Also I have been having issues with the 24GB GPU overheating as it's a P40 without any self cooling and I had to improvise
a method to try and help cool it down. But the constant use of the FLUX 20/25 step model is causing it to overheat and stay
overheated causing the Flux generation to take 7-9 minutes instead of 2-3. So I have also removed the Flux 20/25 step model
until I can get enough donations to purchase a better card with cooling that can produce a result in a much shorter time.
- New update
I have fixed the issue with Unicode characters causing the generation to not start.
I have also fixed the issue with saving Unicode characters in the image.
Feel free to use Unicode characters if you want from now on.
-New update
I have fixed the issue with the search engines not always updating the search results.
New update:
- Fixed issue with Song Player sending too many bytes when sending a requested song. Only about 1 second to load a song now.
- Cleaned up logging of song requests.
- Cosmetic changes to player like removing brackets and file extension from songs displayed, bolded text to read easier.
- Added song name, play/pause button and seek bar below volume bar for current song playing.
Click the purple icon in the top left to open the Song Player.
In case you were not aware, you can drag and drop your own MP3's or OGG's into the Group Chat and they will get added to the Song Player.
The Song Player currently has 269 song selections. See Rules/Help for more info on the Song Player.
You can also drag and drop images into the Group Chat to share with others. PNG/JPG/GIF/WEBP
New Update
- Added a Search feature to the Song Player to make it easier to search songs or artists.
- Fixed Remember function which somehow broke.
- Added Auto scroll to SHUFFLE feature and removed the global song display.
If you only want to shuffle certain songs, you can click the checkbox next to each song you want to shuffle and the
shuffle will be limited to those songs. If you click remember when some checkboxes are checked then it will SAVE
your selection. Next time you start the player (With no boxes checked as any checked boxes causes clicking
remember to SAVE) click remember to automatically bring back your previously saved selection.
New Update
I got sick and tired of the slowness from running the website under IIS 10 and pretty much got everything fixed so I decided
to upgrade the website to a better webserver. For about the last 4 days I have been working on converting over to Nginx.
It's a lot of work converting everything. I have almost 200 files in the root directory and several folders to go through.
The directories changed for everything so it's a major undertaking to switch but the website runs faster with the upgrade.
What I have fixed so far:
- SDXL generation with image auto populating. (All models and Loras)
- SD3.5 generation with image auto populating. (Turbo TEXT/NO TEXT and TurboX options)
- Did 3 days of work alone on the Song Player before switching. If you have not checked it out, you should. Top left music note.
- I also got bored before deciding to switch webservers and have been working on a video player.
Not sure what I'm going to use it for yet, but it works great. If anybody wants to try it out give me a shout.
I could use more people for testing.
Still to do (That I can think of...):
- Flux generation (All models)
- Lumina generation
- Fix the buttons for Face Fix, Upscaler, and Remove Background. Also add the same options to the NSFW category.
- Fix the GPU percentage indicators. Seems Nginx has a different way of handling web sockets so it's been a pain to get working again!
- And after all that, I have to do it for all the other languages.
Update: Fixed the percentage and Loading... indicators when generating.
New update:
All the image generation models have been fixed. The only thing still broken are the buttons for face fix, upscale, and remove background.
But these were actually broken before the upgrade. Don't know how the heck it happened. I have a ton of
backups but during one of the major conversions using the AI models they broke these features somehow and I have been
having a hard time getting a working code solution. It really messed it up good, but there were so many good changes I
could not revert back. For some reason the remove background no longer works on the selected image, it always goes to the
last image generated and I have been a hell of a time trying to fix it. You would think it would be an easy fix but I have
tried using the top 3 AI programming bots multiple times and they always screw it up. I will keep working on it though
and hopefully will figure it out shortly and also add the same features to the NSFW folder.
New update:
Realized switching to Nginx also broke the search engines. They have now been fixed.
Fixed a configuration issue.
Removed ads from search engine pages.
New update:
- Added outline to categories and made the categories nicer.
- Added nice purple and gold background to categories when hovered or clicked.
- Categories will now stay highlighted when clicked in case you forget where you are.
- Got rid of the plain dropdowns and textboxes and added a futuristic look to both.
Completed for English, Spanish, and French languages.
I got bored and decided to start working on a web browser-based game.
See if you can beat my current high score... It's addictive!
Link: https://aiimagecentral.com/robot-invasion.html
Made another game in the series... Robot Invasion II
Check it out at https://aiimagecentral.com/robot-invasion-II.html
New update:
Finished Robot Invasion I and II
- Added arcade page for showcasing games Link: https://aiimagecentral.com/arcade.html
- Added extra life at each 50K or 30K
- Added High Score that increases matching your score if you beat the high score. Initials switched to YOU.
- Added Enter initials at end of game for high score
- Increased number of levels to 50
Working on another game called Robot Wars. Think Mission Impossible mixed with Gauntlet and Robotron :)
Early screenshot... Got the moving, shooting, robots attacking, scoring, some sound FX, level designer, and levels working.
New update:
Been working on another game called Robot Wars that I think is going to be really good when finished but it's going to take awhile.
The concept is to be a mix of Gauntlet, Mission Impossible, Robotron, and probably some elements of Might & Magic.
But I am also thinking of building it out to be a whole bunch of games, puzzles, and challenges creating a whole dystopian world
where the robots have taken over and the goals of you and the resistance are to take back control of the planet.
Imagine completing missions where the missions take you into an entirely different games in the Robot Wars universe.
Maybe some games will be multiplayer even.
Right now, it's still early in development and just started working on the Robot Wars game in the last 3 days.
I have since the last message got all the functions so far designed and working 100%.
Added an inventory to pick up items but the items don't do anything yet except for the extra life one.
Also need to create 32x32 images for each item still and decide what items are available.
Created a system where some walls are solid, some can be destroyed, and some others can secretly hide objects to discover.
I still have to figure out which elements from which games I am going to incorporate as there is a lot of different ideas in
all of those games mentioned. Like gauntlet has health you have to constantly replenish, or some games have timers and then
some have lives.
I may do a combination of 2 of those and am thinking I really want to have character building with health.
I like the idea of building a stronger character making it easier to complete game missions.
I have created 10 testing levels to play around with the game mechanics so there is something there that is playable, but
I plan on adding so much more that this will not do it justice.
My friend was helping me test and said it was too hard so I found a use for my video player and added it to the website and c
reated some videos to show how to beat each level. Try first before looking to see if you can figure them out. They are not that
hard yet and not final levels like I said I'm just using for testing game elements right now.
Links have been created for the video and arcade sections. Check them out.
It gets tedious trying to make the levels inside the game source code so I decided to make a level editor.
Then I thought why have to rewrite every new game from scratch so I decided to turn it into a game maker system
and created a game engine that can run the game_data.json files.
So I have been working on that for the last 5 days or so and it's coming along nicely.
Most of the stuff is added to both the level editor and game engine to make it functional.
It's modular so in the game settings you just pick what features you want the game to have
and they automatically get added to the game when run in the game engine.
Things completed:
- Everything not listed below. :)
Things I know I still need to add...
- Readd Pause button AI removed! COMPLETED
- Biggest thing I still need to do that I don't have added is animating the objects for walking and stuff
- Enemy movement system COMPLETED
- High score feature (Already have code)
- Explosions (Already have code) COMPLETED
- Bonus timer feature
- Timer countdown feature
Over time I will add new features to support building other types of games.. for example (ghost chasing from mrs. pacman)
I am also thinking I may make a section with the editor and game engine online so people can create their own games
and then upload the files and anybody can play it. Let me know what you guys think.
Here is a screenshot of the editor:
The game engine is coming along nicely. I created a demo video of a gauntlet type clone for testing.
Still working on the Game Engine. Working pretty good now for Gauntlet/Space Invaders type games.
Still a bunch of stuff to add to the Game Engine but the basic gameplay works well.
Created new links for two games I have been working on with the engine for testing.
Fixed screen games grid size is now dictated by the maker as they don't work well when changing grid size.
For scrolling games the grid size can be selected from the menu.
Gauntlet Clone - https://aiimagecentral.com/gauntlet.html
Robot Invaders - https://aiimagecentral.com/invaders.html
CONTROLS: Cursor ←↑↓→ to Move (Hold Two keys for Diagonal) | SHIFT + Cursor to Run
CTRL or Z + Cursor Key to Pick Direction and Shoot
SPACEBAR to Start/Restart | P to Pause
They will get better over time as I get the Game Engine working for everything how I want.
Still working on the Game Engine/Editor...
Latest version screenshot:
New update:
- Still working on the Game engine and level editor. Added rotation of the enemy and player objects when changing direction.
- Added 8-way images as well for direction but ctx rotation is smooth, fast, and only needs 1 image. But rotation only works for certain movements so need multiple ways.
Have to add options for movement methods.. like locked on x/y axis like space invaders for example.. Option for the game to automatically fire in the direction the player
is pointing in, option to use 8-way images which I removed when I added the rotation method. Still need to add shoot options as well but have those already in the game
engine, just need to add to editor. Lots of changes to the editor like background color of level playfield now matches selected background color for level. Added 8-way
images for 8-way direction. Added adjusting grid line colors for on new background feature. Added true color support. Also need to add options for enemies regarding
shoot (More options.. fire rate, speed, distance, etc) added to editor.
- Switching to base64 true color images has sped up the engine like crazy. Can now play even in full screen without any noticeable slowdown.
Even with hundreds of enemies on the screen moving. It's pretty awesome. :)
- Fixed the object rotation, also implemented the best algo for it - Lanczos resampling to handle 45 and 90 degree rotations although I can set it to any degree.
Fixed the game settings so the loaded values actually populate in the game settings instead of the defaults.
If you want to check out the latest editor, you can here: https://aiimagecentral.com/level-editor7b.html
I just added a tab menu for player and enemy settings. Needed more space for all the options!
I am still working on it and still a bunch of stuff to add and work on. Like only grid and chase enemy formations I have actually finished. Well still probably add more options
but the enemy moves as expected. But you can make fully functional games. Just I would need to get the file you create and setup a game engine to play that file for you
to play it. I will probably setup sections for building your own games and other sections for people to play them.
You can try the latest game engine here: https://aiimagecentral.com/game-engine-v1.97p.html
Here is an image of the editor ATM:
Added some new settings to player settings. Can now control movement and shooting direction. Also added a feature
that allows the images to rotate based on direction moving (For top down view games)
Bunch of new stuff to add still..
Need to reorganize the object structure to accommodate more objects. Thinking...
walls (unbreakable, breakable, hidden) 1-299, items (pick up) 300-599, items (player/enemy ammo images) 600-649,
items (weapon images) 650-699, enemies (all) 700-949, players (all) 950-999
-Add scroll speed to settings.. need direction as well as speed to scroll. Needed for games that automatically scroll like scramble.
-Add object fields to allow several things.. co-ordinates offsets , number of shots to destroy, hidden items, ammo qty, rotation and other simple movement animations.
-Add 8 way images option to use them as animation or something like wall destruction shown after each hit for multi hit destruction
It's crazy how many options you need for stuff to allow variety. The above should be the next changes.
Here is an image with the new player settings.. BTW all these settings are modified on a level by level basis so you can have completely different game types for each level.
Fixed a bug in Robot Invasion that was causing one of the enemies to get really large blocking the other enemies starting at level 5.
Fixed another bug in Robot Invasion causing the play speed to be greatly increased.
Also fixed so that pressing Space restarts the game so don't press it if your doing good. lol
Also applied same fix to Robot Invasion II so pressing space anytime restarts the game.
Been working on converting the website to partly work in Windows 10 and partly work in WSL 2 Ubuntu.
Ubuntu can handle many more connections and is lightning fast compared to Windows.
There is a lot to change but I have already converted a lot over.. Image generation and all the functionality for that works.
Going through categories and all the other buttons on the webpages are working except the following.
Still need to fix: Remove image button, Face Fix button, Upscale button, and Remove background button.
Still need to incorporate SSL certificate (Once everything is working 100%)
Need to fix issue with Info not always showing when clicking the file info button. Multiple formats, not always displaying.
I think all the rest is working. But still needs to be done in stages. Once everything is working and I have converted the website
part to Ubuntu completely then I can remove some of the redundancy created by the two systems.
The node.js server (Handles chat, API calls, Video, and Song Player) and the API's and image generation are all staying on the Windows side.
All website interaction (PHP/HTML5/Javascript) will be through Ubuntu.
It's already blazing fast to load any page now in the new system.
Still working on converting the website over to WSL 2 Ubuntu...
Fixed Remove Image buttons.
Fixed INFO button not showing all stored comment formats.
Converted, fixed, and added searches for all languages for both prompt search and image search.
Converted All the files for the Spanish language. Added TensorX model and arcade/video links
Converted All the files for the French language. Added Fastflux model and arcade/video links
Four more languages to go.. (16 more index.php files)
Then I can add SSL and go live with faster version.
Looks like converting over to WSL 2 version of the website is going to be delayed.
Seems I need to upgrade to Windows 11 in order to get the WSL 2 mirroring feature needed to make the website public on the internet.
In the meantime I converted all the improved versions from that install by fixing the paths back to work in Windows.
I will need to purchase a new computer in order to upgrade to Windows 11 as this computer does not have a TPM module.
I don't have money for this so it may take awhile. But the Windows version is now improved and speed is not bad.
It's like 1-4 seconds for the Windows version vs like .25 to .5 seconds for the Linux version which can also handle way more
connections (Concurrent connections is 50000 for the Linux version). Windows OS can't even come close to this.
Added the HiDream Fast 16 step model to the list of available models under the English language.
This model is quite slow and even the 'fast' version with 16 steps takes 2-3 minutes.
It takes too much memory to run in fp16 so I have to limit to fp8.
This model is interesting in that it uses llama 3.1 8b instruct on the prompt.
Been pretty much out of commission the last 3 weeks due to some personal health issues.
Still need to make the last changes mentioned to the Game Engine.
Robot Wars has pretty much been put on hold since I plan on making the final game with the Game Engine.
Still not sure on all the game elements I want to add to robot Wars but still want to make it like an expandable
universe where you play multiple games over time to advance in the entire game universe.
I have been having issues with the image generation over the last couple of months.. with the generation lagging.. getting stuck
and just taking way longer than it should. I was reading some stuff on Reddit, and it may be from the newer Nvidia drivers so
I have reverted back to 531.61 as after that is when the issues with SD generation taking longer start.
Seems the default settings for Comfy UI no longer work well for older GPUs so I have added some new flags on startup.
Issue for anyone interested: https://github.com/comfyanonymous/ComfyUI/discussions/4457
See if this helps...
I finally got the remove background working again properly. Option has been added to all English model pages.
(Click on thumbnail image to see Remove BG button)
I also now fixed it so that you can remove the background on any existing images in any of the categories.
For NSFW, the new image will populate in the NSFW folder. For all other categories, the new image will populate
in the Requested category.
Also, the PNG comment is now displayed when the INFO button is clicked on an image where the background is removed.
This will give you the path/category to the original file with the background intact.
Added the HiDream model to all the other languages.
Added the Remove BG button also to all other languages.
Still had issues with generation speed so I started playing around with previous settings (mainly removing command line options).
I think it's now back to being as fast as before. However, the 24gb GPU is fanless so the longer generations like HiDream, Lumina,
and 20-25 step SD3.5 & 20-25 step Flux are still pretty slow. The card throttles down to around half speed once it hits the temperature limit
which it will about halfway through the 2–3-minute generation. The load times when loading a new model have definitely been improved
as well as the iterations/second.
Renewed the Domain and SSL certificate for another year. SSL is such a pain in the ass...
Also needed to update the chat server to work with the .pem certificates instead of the .pfx certificate as my new one is password protected
and the previous .pfx file did not have a password set causing an error when attempting to run.
I have been working a little on the game editor and game engine. Adding a few new features and fixing some bugs.
- Added 3 second invulnerability on new life when reset type is player and not level. Not needed for level reset.
- Fixed enemy kills while player invulnerable is active.
- Added level auto scroll setting for games like Scramble to game editor and game engine.
- Fixed inventory display.
- Fixed run enabled option under player options - Added check to fix running speed when rE not true.
- Made a better Help menu using HTML instead of an alert.
- Fixed issue with the game editor not saving the object Name changes when applied.
- Added 8x8 grid size option to 16,32,64,96 preset object sizes
Created a new forum for the Game Editor and the Game Engine.
I have been working on a new interface that allows you to customize the size of the image.
So far I have only implemented it for the Stable Diffusion 3.5 models.
Let me know how you like it...
Maximum size is 1280x768/768x1280 or 960x544/544x960 depending on model with the option to upscale x2.
Stable Diffusion 3.5 models do not work well for images below 600x600.
Added custom size (X/Y sliders) option to Flux models.
All the Flux models work well with smaller images even down to 100x100
Added custom size (X/Y sliders) option to SDXL.
All SDXL models work well with smaller images as low as 250x250.
Added custom size (X/Y sliders) option to Lumina 2.
Lumina 2 does not work well with images below 500x500.
I am removing the HiDream model. The quality is not anything special and it's ridiculously slow to load for some reason.
I blame the ComfyUI API but unfortunately there are no other options with this kind of model support.
A1111, Forge, and InvokeAI all useless.
I had set the search engines to use AI generated descriptions but I don't think this was the best approach as the AI generated
descriptions seem kind of bland due to issues like recognizing famous people. So now I have instead created a python script to
extract the actual prompts from the images to be used by the search engines (Same as prompt under INFO).
I have also fixed the ANY and ALL search engines so that the order of the words does not affect the results.
Added a new feature to all models that lets you choose the shape of the image of either a rectangle (Original) or a square.
For SD3.5 - Up to 1280x768 rectangle or 1024x1024 square
For Flux - Up to 960x544 rectangle or 768x768 square
For SDXL - Up to 1280x768 rectangle or 1024x1024 square
Just click the image of the rectangle or image of the square to switch.
Size matters. Models have limits on how small an image they can produce without artifacts.
Stable Diffusion 3.5 models do not work well for images below 400x400 (Even with CFG set to 1.5).
All the Flux models work well with smaller images even down to 100x100.
All SDXL models work well with smaller images as low as 250x250.
Added new square and slider options to all models in Spanish and French now as well.
Still need to convert HI, ZH, and AR (Hindi, Chinese Traditional, and Arabic)
Further enhanced the image search engines to give better results. The searches will no longer match partial words preventing
excessive erroneous results by placing a space at the end of the word(s) to prevent partial matches like 'cat' matching 'catapult'.
However, adding the space after each word prevents other important plural matches. IE. You would expect
searching for dog would also bring up images of more than one dog.
So I have also added checks for all words with 's' and 'es' added to the word(s) in order to catch more images.
I believe this the best way to make a useable search engine without getting super technical with dictionaries.
Working on changing the interface. Cleaner layout.
You can check out the new interface from these links:
SDXL - https://aiimagecentral.com/index-en4.php?dir=1080p%2FRequested
FLUX - https://aiimagecentral.com/index-FLUX-en4.php?dir=1080p%2FRequested
SD 3.5 - https://aiimagecentral.com/index-SD3.5-en4.php?dir=1080p%2FRequested
I have not changed any of the other page links, so clicking on any other links will bring up the old pages.
Finished the menu to switch between SDXL, FLUX, and SD3.5.
Got rid of Lumina 2 model. Too slow like HiDream.
- Added Page # and next/prev buttons to top of categories now under the new format.
- Added Flux Fill to Flux menu.
Still need to create the model info panels for each model. Will give information about the selected model.
- Converted all the other tabs to new format (prompt search, image search, arcade, about, and rules/help).
Still need to create the model info cards... Still debating what info to put on them and best method to create them.
Also converted Flux Fill page. All webpages converted to new format!
- Added new auto scroll feature so that when you select a category or use the next/prev buttons, it scrolls down to show the images.
- Changed the graphics on all the buttons to look nicer and get rid of the washed-out look.
- Added left/right cursor key controls to switch between different images when in full screen overlay mode.
You can press 'ESC' to close the full screen overlay or click somewhere on the image to close.
I have increased the size of the game playfield for both Robot Invasion and Robot Invasion II.
I also fixed the pause button 'P' so that everything pauses. Just press 'P' to Pause/Un pause.
Just realized I need to change the top menu as it still has the old format.
- Converted the top menu for Robot Invasion & Robot Invasion II to the new format.
- Added a surprise ending if you manage to beat level 50 in Robot Invasion I.
- Increased Robot Invasion II number of levels to 100 (Was 50, I had 6 lives left after beating level 50 so increased).
- Added a surprise ending if you manage to beat level 100 in Robot Invasion II.
Can you make it to level 50 in Robot Invasion I and stop the first invasion?
Can you make it to level 100 in Robot Invasion II and stop them again?
I did not realize in all the changes I broke the Flux Fill model. This has now been fixed.
If you are not familiar with this model, it is an image-2-image model used to make changes to an existing image.
Drag and drop the image you want to modify to the selected area. Mask the area of the image you wish to change.
Then type in a prompt to describe the modification you are trying to make and submit.
For example, to change some one's eye color, mask the eyes and type a prompt like red eyes.
- Limited to 960x544/544x960 maximum size as it already takes like 4 minutes even at this size.
- Flux Fill CANNOT DO NUDITY so don't waste your time trying.
Latest Update:
- I added the background back to the top menu bar for the website.
- Redid the graphics for the Model Menu to fix spacing issues.
- Did multiple HTML fixes making sure all the webpages match the new format (Navbar font, spacing, etc.).
- Made some changes to the Flux Fill webpage. Minor changes to page format and some graphics changes.
I have updated the model interface for each model to limit the minimum resolution for each model based on how
well the model works with lower resolutions. In testing I have found if you go below a certain resolution then
the image generation is flawed.
The limits are set as follows:
Flux models work the best with smaller sizes and allow you to go down as low as 100x100.
SDXL work the next best with smaller resolutions as low as 250x250.
SD3.5 models are terrible with smaller sizes and only work well from about 600x600.
I added some old school arcade games to the arcade section.
You can now play my own creations as well as Galaga, Pac-Man, Frogger, MS Pac-Man, Tempest, Donkey Kong,
Dig Dug, Asteroids, Centipede, Joust, Rampage, and Rampage World Tour.
Cursor keys to move and CTRL key to fire. SHIFT Key is the secondary button.
SHIFT key in Tempest to use the Super Zapper.
SHIFT key in Asteroids for thrust.
SHIFT key in Rampage to jump. CTRL key to punch.
SHIFT key in The Simpsons to jump.
SHIFT key in Gauntlet to use potion.
SHIFT key in Out Run is brake. Z key is LOW gear; X key is HIGH gear. CTRL key to accelerate.
SHIFT key in 1942/1943 to do barrel roll. CTRL to shoot.
SHIFT key in In The Hunt to shoot up and drop mines, CTRL shoots left to right.
SHIFT key in Contra to jump. CTRL to shoot.
SHIFT key in Street Fighter 2 to punch, CTRL key second punch, Z key for third punch, X key for leg kick.
SHIFT key in Mortal Kombat/MK 3 to kick, CTRL key to punch, Z key second punch, X key second kick.
SHIFT key in Omega Race for thrust. CTRL key to shoot.
Games using more buttons may also be using Z/X/C keys.
Tempest, Donkey Kong, Dig Dug, Asteroids, Centipede, Joust, Rampage, and Rampage World Tour are using the
EmulatorJS emulator.
To change the keyboard controls, click on the game controller icon in the bottom left.
Added 26 more classic arcade games for a total of 38.
I have added the best arcade games of all time that work well with a keyboard or gamepad. Arcade games
made available from the Internet Archive's Arcade section. Link: https://archive.org/details/internetarcade
Some games don't work well like Spy Hunter when trying to use the keyboard and are not added.
Firefox does not display the screen graphics at the right scale and it's very small unless you go to full screen
so I don't recommend using Firefox. Chrome and Edge both work fine displaying at the intended resolution.
Added a small sign on the Arcade page showing the arcade controls.
Added 6 more classic arcade games to the Arcade section. I think that is the last of the 80's arcade
games that work well with a keyboard and people will remember.
- New update
- Converted the all the French pages to the new format.
- Changed the default page for the website to automatically switch to user's default language for the webpage.
- Removed language switching from individual pages. Now language is only detected from https://aiimagecentral.com
New Update
- Converted all the Spanish webpages to the new format.
- Modified the main webpage landing pad to auto switch to the user's default language if available (ES,FR,ZH,AR,HI)
- Got rid of the auto language switching on all other webpages.
- Converted all the Hindi webpages to the new format.
- Converted all Arabic webpages to the new format.
- Converted all Chinese webpages to the new format. Switched to simplified Chinese.
- Fixed all the search engines to work with any language. Language will be converted to English to perform
search since the databases are in English. Previous code had issues with symbol-based languages. My filter
was removing the symbols. lol
- Fixed all the button graphics for all the languages. Buttons not sized right and different patterns over time.
- Removed the model info graphic since I have not figured out what to do with it yet or if I want to keep it.
- Added the Song player back to the website. It's now on the Arcade page. Changed to automatically start shuffling songs.
- Created a Reddit website several weeks ago. Can access from here: https://www.reddit.com/r/AI_Image_Central/
- Fixed issue with sliders not working correctly when slid all the way to the minimum value. It would show the correct value (For example 300x300 for X/Y) but was sending the value '1' for X in error causing the generation to fail.
New update
- Modified the progress bars to now display whichever model is currently loaded.
New Update
- Modified the progress bars to now display whichever model is currently loaded for the Spanish webpages.
- Added a new model called Artius to the SD3.5 model menu. Looks pretty good even with 4 steps in my short testing.
- Added the new model to the Spanish webpages as well.
- Also moved the default SD3.5 model to the [NO TEXT] version since this model is fp16 and has better quality. Users should read the Help section in regard to adding text.
Let me know if you encounter any issues.
New Update
- Added model load and the new model Artius to the French language.
- Added model load and new model Artius to the Chinese Simplified language.
New update:
- Added a new model called Z-Image to the Flux page. The model uses Qwen 3.4b for understanding and is quite good.
- Modified the Flux page so that the sampler, scheduler, and max resolution are now changed when you select a model
allowing me to customize these settings per model.
Recommend using 9 steps as loss of detail is noticeable at 4 (Big loss) and 6 steps. Best open-source model to date for most tasks. Recommend Flux Dev Turbo as second-best open-source model. Third choice is SD3.5 Turbo. If one model doesn't work, try another.
New update:
I have added another option to the Z-Image model allowing you to do full 1080p images. This takes 2-3 minutes due to the high resolution. You can also click the Upscale X2 option to increase to 3840x2160.
New update:
I have now updated all the other languages to the latest format with the Z-Image model.
I also updated my games so they also display in the user's language now.
Just a few small fixes left for the Arabic versions of the rules, about, and searches.
New update:
Added a link in the menu bar for the Reddit sister site.
I have modified the Z-Image model to use a better scheduler and increased the 8 step option to 9 steps.
I have also added a 6 step option to the Z-Image 1080p section.
New update:
- Added auto language translation to English for the Flux Fill model for all languages.
Remember for Flux Fill: Mask the area of the image you want to change only. Prompt what you want changed in the masked area.
- Modified the Z-Image dropdown to reverse the order of steps. The default is now 9 steps as 4 and 6 steps lose a lot of detail.
It also works much better when adding text with the higher step count. Unlike other models, the loss of detail is noticeable.
In most use cases, this is the best open-source model available.
- Small fixes to slider controls
New update:
Some model's dimensions are hard coded to use values in multiples causing undesired resolutions.
For Flux models, trying to get 960x540 does not work as it has to be 960x544 as well as 1920x1080 has to be 1920x1088.
I have modified the code so this is now hidden behind the scenes and the images will be trimmed to the correct size.
960x540 is important because upscaling it x2 gives you 1920x1080 (1080P).
For Z-Image, you can also upscale x2 from 1920x1080 giving you the 3840x2160 (4K) resolution.
Looks like my 24GB GPU is fried. Since no one is donating I doubt I will run the site much longer as I don't have the money to replace the GPU. I moved all the SDXL models to the working GPU so all the models are back to working.
Will leave up for now. If you want the site to stay, consider donating. Link on rules page.
Site was down last week due to issues with ISP. Issue finally resolved.