Skip to main content ↑ ↓
Topic: Welcome to AI Image Central! (Read 11891 times) previous topic - next topic

Re: Welcome to AI Image Central!

Reply #105 –
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.
 

Re: Welcome to AI Image Central!

Reply #106 –
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.


Re: Welcome to AI Image Central!

Reply #107 –
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.

Re: Welcome to AI Image Central!

Reply #108 –
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)

Re: Welcome to AI Image Central!

Reply #109 –
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.
 



 

Re: Welcome to AI Image Central!

Reply #110 –
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.
 

Re: Welcome to AI Image Central!

Reply #111 –
- 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.


Re: Welcome to AI Image Central!

Reply #112 –
- 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!

Re: Welcome to AI Image Central!

Reply #113 –
- 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.

Re: Welcome to AI Image Central!

Reply #114 –
- 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.

Re: Welcome to AI Image Central!

Reply #115 –
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.

Re: Welcome to AI Image Central!

Reply #116 –
- 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?

Re: Welcome to AI Image Central!

Reply #117 –
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.

Re: Welcome to AI Image Central!

Reply #118 –
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.

Re: Welcome to AI Image Central!

Reply #119 –
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.
 
AI Image Central - Welcome to our forum!