STOP! THESE MODELS ARE NOT FOR PROMPTING/IMAGE GENERATION
These are the new ControlNet 1.1 models required for the ControlNet extension, converted to Safetensor and "pruned" to extract the ControlNet neural network.
Also Note: There are associated .yaml files for each of these models now. Place them alongside the models in the models folder - making sure they have the same name as the models!
These models are embedded with the neural network data required to make ControlNet function, they will not produce good images unless they are used with ControlNet.
To be honest, there isn't much difference between these and the OG ControlNet V1's. These models are further trained ControlNet 1.0 models, with an additional 200 GPU hours on an A100 80G.
control_v11p_sd15_seg
control_v11p_sd15_softedge
control_v11p_sd15_normalbae
control_v11p_sd15_scribble
control_v11p_sd15_mlsd
control_v11p_sd15_openpose
control_v11p_sd15_depth
control_v11p_sd15_canny
control_v11e_sd15_ip2p
control_v11e_sd15_shuffle
control_v11p_sd15_inpaint
control_v11p_sd15_lineart
control_v11p_sd15s2_lineart_anime
control_v11u_sd15_tile
Download these models and place them in the \stable-diffusion-webui\extensions\sd-webui-controlnet\models directory.
Please consider joining my Patreon! Advanced SD tutorials, settings explanations, adult-art, from a female content creator (me!) patreon.com/theally - I also have a write-up of ControlNet and will be updating with the latest news/developments!
Description
FAQ
Comments (49)
Thank you for taking the time to get these updated models added. I'll look into the issue with the ability to add yaml files to ControlNet types of models
Looks like we already got it addressed and you've uploaded them. It appears the notice isn't showing though, so we'll get that added.
what model did you use on the first picture? it's look awesome
@nicetry20010 Hey, all of these are from my next Ally's Mix - this will be the fourth one. It's geared towards realism but still retains the feel of my work! I really think this one is going to be special!
What's the total download size? Mind adding that to the description? I blew out my SSD earlier today with the unpruned ones lol.
may i know what kind of prompt you used to get that painterly looks?
Do these work with the existing ControlNet plugin for Automatic1111? Mine is showing the latest as being 6dd043aa (Mon Mar 6 16:59:46 2023).
It might be good to also mention that these models are not meant for merging into models for generating images, since some people refuse to understand this
I don't think anything I could say will convince them :)
Will there be also a "Difference" version of this?
Unlikely! There was so little variation with the last difference extracts, I'm not sure it's worth it!
@theally I see, thanks for the answer!
@theally controlnetV1.1 has the lineart models which was not present on V1. So those alone is worth the download
@spoffninja I extracted the original ControlNet models two ways - one was a "difference" extract by calculating the difference between ControlNet and SD 1.5, the other was prune/fp16 - that's what @Albedowo was referring to - I wasn't saying there are no differences between 1.1 and the OGs :)
i saw that you have to do a different install for this becuase its not the main control net...is there an eta when controlnet will be bumped to 1.1 officially?
You can use (some of) these with the existing ControlNet extension, but no, no idea when it will officially support all these models. Hopefully soon!
@theally excited to use them but the set up is rough for me
Thanks a lot for your work! :D
Can we just delete the old models and use just these new ones? Or should we keep them and use both old and new ones?
Thank you for sharing those ConrtolNet models.
Is the depth model included here based on the original upload of 1.1 or the bugfixed version? Per the original github repo:
2023/04/14: 72 hours ago we uploaded a wrong model "control_v11p_sd15_depth" by mistake. That model is an intermediate checkpoint during the training. That model is not converged and may cause distortion in results. We uploaded the correct depth model as "control_v11f1p_sd15_depth". The "f1" means bug fix 1. The incorrect model is removed. Sorry for the inconvenience."
It's fixed! The model here is the control_v11f1p_sd15_depth model.
what did you use to convert the models from pth to safetensors? Im asking because this models get updated often and would like to do it myself.
I am getting A LOT of "Model not found" on these... some of them work, some of them just utterly fail. Any ideas / suggestions?
Do you get the same with the official models? I'd be extremely surprised if not. When do you get the "model not found" error? Do you have the .yaml in the same directory, with the exact same name as the model? And you're selecting the appropriate preprocessor?
Well for now they seem to be (at least somewhat) working. Haven't tested everything so far. Also, I should have been more clear- I wasn't meaning to imply there was anything wrong with the models themselves. Yes, the YAML models, which I got from here on the model page, are in the same folder as the models .
Mostly, I was just looking for help. Anyway, it only seemed to bean issue with certain models. I haven't tried the "official" ones yet. Usually I get the error when I start to generate something with that model. It will make SOMETHING because it defaults to standard SD. I am on Mac M1 and some like Segmentation (I don't know if it's just the pre-processor or the model or both) just give me the finger and utterly refuse to work. that's not your fault as the same happened with 1.0 models and they weren't even yours. This is just the price to pay, I guess, for living on the "bleeding edge" as the saying goes, of technology.
Appreciate all your hard work providing these resources.
Thank you so much for providing these! They take up so much less space and work just as well.
tested a few of my creations these models vs full, same seed same result very happy just like the 1.0 models. Huge thanks.
What's the purpose of the inpaint model when we can already inpaint?
Yeah, I have zero idea!
@theally lol ok thanks anyway for uploading it!
@icingqueen perhaps it enables you to use any base model for inpainting? so that you don't have to beg the author for a dedicated inpaint-checkpoint?
Controlnet's inpaint can actually outpaint as well using lama or harmonious global, if you have a square image and then in img2img you specify 16:9 ratio and bigger area and select controlnets inpaint+lama, then it will outpaint. There's some tutorials on it if you want more detail.
Can you please upload the fixed tile model? thank you :D
Took me a while, but it's here now!
Someone realize how to use Tile model? When I add control with tile in img2img and use SD upscaler – SD generate same image in all of the tiles...
In Reddit there's many examples how to use it even Olivo Sarikas explained how to Upscale using tile model
I forgot that the current version of tile model here doesn't work well you need to download the update from huggingface
@Blackhat98 You can't use SD Upscaler, but you can use the Ultimate SD Upscale extension. You can also combine Tile with "my prompt is more important" to do arbitrary prompt-based transfer (which is what I'm showing in the example image).
You can also use Tiled Diffusion instead of the Ultimate SD Upscale, though I have not had such good results, but I have seen others who have. Just need to play with it more.
https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111#-combine-with-controlnet-v11-tile-model
Please update of the tile model to the version "e" instead of the "u", control_v11e_sd15_tile
Done!
awesome job, please update to the version e
Can some Hero update the 'tile' model with the recent update and in .safetensor?
TY!
Took me a while but I finally got around to it! FP16/Safetensor uploaded!
Details
Files
controlnet11Models_shuffle.safetensors
Mirrors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
controlnet11Models_shuffle.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
controlnet11Models_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
controlnet11Models_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
Shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
shuffle.safetensors
shuffle_cn11.safetensors
controlnet11Models_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
controlnet11Models_shuffle.safetensors
controlnet11Models_shuffle.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
control_v11e_sd15_shuffle_fp16.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
