Stable Diffusion checkpoint dropdown menu

[Hamguy21]

The Dropdown menu is very confusing when you have several .safetensors models, allowing custom images or icons would make this more helpful.
also a separator or group feature would be nice.

[Woisek]

What exactly is confusing with an alphabetical sorted list? What Icons do you want to use to improve this? You can safe the checkpoints in groups on OS level and the drop down is also a search/filter field ...

[Hamguy21]

a bunch of text doesn't give you an idea of what exactly you are loading when every model has a different name and different use case.

For example the Lora or checkpoints gives you a list of image cards you can add a .jpg or .png file too, I know exactly what i am loading because I have an image specific to that lora.

[Woisek]

a bunch of text doesn't give you an idea of what exactly you are loading when every model has a different name and different use case.

Normally, one should know what he is doing and what exactly he wants to use, but okay ... Maybe when in search for something this is different, but ...

For example the Lora or checkpoints gives you a list of image cards you can add a .jpg or .png file too, I know exactly what i am loading because I have an image specific to that lora.

... so you obviously have exactly what you want. I don't get now what the problem is. You have a dropdown with a searchable/filterable list of checkpoints and you have a 'gallery' with preview images of the checkpoints the other stuff (given you have all the preview image set). Just choose what you like the most. ¯_(ツ)_/¯

What now is confusing or problematic? 🤔

[ghost]

The naming and lack of thumbs is problematic, the pull-down is the top most UI element, amd a thumbnail next to a proper name of the checkpoint is standard affair on every AI image gen site atm. Why not this? Seems like a missed opportunity that IS noticed by many

[noisefloordev]

The Dropdown menu is very confusing when you have several .safetensors models, allowing custom images or icons would make this more helpful.
also a separator or group feature would be nice.

You can do this:

• Put an image next to the model with a matching filename (abcd.safetensors -> abcd.jpg). Then instead of using the checkpoint dropdown, open the "Checkpoints" tab and it'll show the images. It works the same as thumbnails for LORAs and other models.
• You can group checkpoints by putting them in subdirectories. They'll group together in the checkpoints dropdown (though the sorting isn't great), and you can filter by directory in the checkpoints tab.

[Woisek]

You write the same thing as I did, but you vote me down? What's up with you, dude? 😒

[Hamguy21]

yes i've already done this, but i thought it would be useful to have the ability to do it on the list. thanks anyways.

[Woisek]

What for do you need that in a list, that is supposed to be ... a list? 🤪
Sorry, but the desired function is already available, so this is neither useful nor necessary.

[noisefloordev]

You write the same thing as I did, but you vote me down? What's up with you, dude? 😒

All I could see was a half page of "what's so confusing about that, I don't get the problem, you obviously have what you want (confused face emoji) (irritated shrug emoji)" in response to a really simple, straightforward question.

[Woisek]

Then maybe you should read the whole post first and not just half of it and repeat what's already written, before downvoting others.

[ghost]

I 100% agree with this, and it's one simple feature that could be added with a settings toggle to disable it if you don't need/want. No reason not to have added this even as a disabled UI option at Lauch and those who appreciate it, we can flip a switch a has zero effect on those who don't need it.

Notice how every single AI image gen site has previews of the models next to the names? That's because if you have little experience in general, have no idea what the checkpoint does or have a decent set that you are trying out, like 10 new models you just loaded, a small thumbnail appearing next to the Checkpoint pull down list can visually show the checkpoints overall theme or style. Does epiCPhotoGasm, MajicMix, Copax Timeless, YamerMIX, WoWXL explain enough? How about wowXLV2_V2.safetensors.... why is the community making progress on things like Card Descriptions, connecting repos and enabling model install in batch? Seems like we need a bit of an overhaul to make this a bit easier for people starting out, the UI is cluttered and fugly, the UI/UX mods created were so nice to see while they were available, just didn't keep up. Being able to resize UI elements? Easily Move things and even add much needed bits and pieces was so nice, SD.Next is working with them to make it standard. just about all others look better, are easie to use, but the extensions and constant support for new stuff makes me love this one. Again, this is why Mage, Dezgo, etc all do this, its actually usefull. Just because you may not need extra UI fluff, doesn't mean we all dont. From my first run of the first thing that I found odd and have been surprised such a simple but helpful and obviously common thing has not been added.

Yes, we have an entire UI tab that sorts and shows previews of models, Lora, etc but that is not nearly as helpful imo. I usually stick on my text2img tab, default ui, and make images, so the ability to click that top element, have quick access to all my Checkpoints and switch quickly is nice, please, add a preview thumbnail to this pulldown, use proper names even, there is literally no reason to argue against this, it would not change a thing for those who shut it off, or, how about disabled by default and only those who want it can enable it? Literally defeats any argument against this feature everyone who uses AI image gen sites seems to appreciate. Why would hosting your own, pre-made site not be worthy of this simple feature?

[Woisek]

So, what you basically say, is, you want a huge content, that is already displayed as a tab (and for a reason so), put into a dropdown field ... ?
Okay ... great idea. We should display every web page in a dropdown field, because ...

Yes, we have an entire UI tab that sorts and shows previews of models, Lora, etc but that is not nearly as helpful imo.

I really want to know, how you think this dropdown should graphically be to be 'more helpful' than the tab where all is listed.
Provide a mockup screenshot, this would be helpful for understanding ...

[ghost]

It's really simple, a small thumbnail image to the right of each checkpoint name in the pull-down, as every popular AI gen site had... it's really that simple to make that drop down provide more useful info to those who think the checkpoint file names are not helpful and switching tabs, unnecessary.

[ghost]

Not sure how a small thumbnail that helps people would hurt? Would that make filtering harder? Would it do anything bit provide a useful detail to people?

[Hamguy21]

also the ability to add metadata to models would be helpful, you could rename them to something human readable and add some extra detail and wouldn't have to replace the name the author gave so you don't lose track.

[Woisek]

It's really simple, a small thumbnail image to the right of each checkpoint name in the pull-down,

Aha ... and how "small" should that be? 100px? 200px? 300px?

as every popular AI gen site had...

Show me one that has that packed into a dropdonw ....

it's really that simple to make that drop down provide more useful info to those who think the checkpoint file names are not helpful and switching tabs, unnecessary.

Are you a troll, or want you just being provocative? The previews in the tabs have all the information you want and need. And what problem do you have with clicking in a tab? The whole UI exists out of tabs.
This is really ridiculous.

Not sure how a small thumbnail that helps people would hurt?

Not necessary, wasting space, wasting resources (double display of images). Plus, it's not here to 'help' it's here to select.

Would that make filtering harder?

Could be, but definitely to scroll through and select. Thinking of having 100+ checkpoints in a dropdown that is then more than 10000 pixels long, is simply ridiculous

Would it do anything bit provide a useful detail to people?

All the details are already there, you just have to use it.

also the ability to add metadata to models would be helpful,

Just click on the hammer&wrench icon and edit away ...

you could rename them to something human readable

That happens on OS level, no need to bother A1111 interna with it.

Seriously, this is a joke, and it's not even April 1st ...

-eoc-