SD 3.5 Support

[a3nima]

What's the status and general possibility of SD 3.5 model support in A1111??

[SunGreen777]

I expected that if a new working version of SD was released, the project would come to life at that moment, we are watching.

[HonZuna]

Good idea, but at the current rate of development, we can't expect it before Christmas. Personally, I don't expect it at all.

[Soulreaver90]

I wouldn't doubt it, but seems like the owner is on a long vacation or break. SD.next already supports 3.5, ontop of Flux and everything else under the sun. The only reason why I haven't jumped is because .next runs a hell of a lot slower for some reason on my AMD card. Owell.

[SunGreen777]

can I get the url for SD.next?

[Soulreaver90]

can I get the url for SD.next?

https://github.com/vladmandic/automatic

[TanvirHafiz]

i tried SDnext, could not get it to work!

[CalculonPrime]

The last commit to this repo was 3 months ago. I haven't tried to use vladmandic's fork but the last commit there was 2 days ago, and it appears to already have 3.5 support.

[TanvirHafiz]

but not easily. i asked, even he said that it has unet and what not. but didnt give me a clear solution.

[digger73]

The last commit to this repo was 3 months ago. I haven't tried to use vladmandic's fork but the last commit there was 2 days ago, and it appears to already have 3.5 support.

Well that isn't correct, last commit was 2 days ago, master is used for releases.

[wkpark]

as already explained above, the dev branch has 33 commits ahead of master, https://github.com/AUTOMATIC1111/stable-diffusion-webui/tree/dev

[FJCCOMMISH]

Still no SD 3.5

[francaleu]

sd3.5? some day? thank you

[jEFFLEZ]

A small idea about control in img2img / SD:

We often rely on a single global strength, but in practice it bundles together very different needs.
Changing the background, adding effects, modifying the outfit, or preserving the face are not the same kind of transformation at all.

I’m wondering whether the better direction would be a per-component strength model, something like:

• identity_strength
• anatomy_strength
• outfit_strength
• background_strength
• effects_strength
• props_strength

Even if the final sampler only accepts one strength, these separate scores could still be used upstream to drive:

• the final prompt
• visual priorities
• masking
• the choice between light or strong img2img
• specialized passes

So the idea is basically to replace one global slider with a finer semantic control layer.

To me, that seems much closer to the actual kinds of transformations we ask these models to perform.

Curious whether anyone here has already experimented with something along these lines.

[fcabanski]

OK? Automatic 1111 is not being changed/updated anymore.