How to install Forge for Stable Diffusion. Automatic1111 but BETTER!

[@sebastiankamph]

Text guide for Patreon subscribers www.patreon.com/posts/how-to-install-98578765

[@IndiaUntold]

It has some missing controlnet features such as IP adaptor V2 etc. So any way to fix that?

[@RiftWarth]

@sebastiankamph Great video on Forge. I'm learning a lot from you. Quick question...is it possible for these AI tools to utilize specific fashion designs, like those found in Chinese apparel, and generate new images depicting individuals dressed in outfits with these patterns? If so, could you please tell me how?

[@RisetOnline]

Hello Seb, Thank you for this video and all your other videos-they've been incredibly helpful to me. However, have you looked into what's been happening with Forge lately? I'm too scared to update Forge because the latest version caused many errors. The author announced that he won't be updating Forge anymore and suggested going back to A1111 🥲 Could you discuss this in detail? My English isn't very good, so I struggle to understand all the latest news about Forge. Thanks!

[@matijaziljak4478]

I just started using stable diffusion out of curiosity a few days ago and my experience with my 4GB of vram wasn't that much fun, but with forge I'm seeing a drastic difference and I can create an image in less than 2 minutes (before was ~11 min). Thank you so much !!

[@Alienfodder]

OMG... this is truly a game changer, at least for me. Even if I'm using a 3060 GTX with 12GB of vram, 4 XL images would take about 15 minutes to generate. Loading the models was also a pain... But now it's all in the past! Thank you, Sebastian!

[@sebastiankamph]

I'm happy that you're happy!

[@gerardoweimann7921]

"Even if i'm using a 3060 GTX with 12GB of vram" while i cry with my Vega 56 xddd

[@nenadm5747]

How to make it work with Amd card? Thank you

[@Bockiiey]

Heard about Forge for the first time today and was hoping for you to post this video :) perfect timing

[@sebastiankamph]

Perfect timing indeed then! :D

[@EskaronVokonen]

In the webui-user.bat file we can reference our A1111 existing folder, so that we don't need to re-download all the models, embeddings, etc. Then play with it, see if we like it better and later decide if we move everything into Forge (and REM back all the A1111 links)

[@G0odGames]

how can we do this?

[@Steamrick]

@@G0odGames set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@EskaronVokonen]

@@G0odGames Like I said (kind of), by editing the webui-user.bat file and removing the @REM (which probably means "remarks", or "comment" if you will) before each meaningful line of code in the webui-user.bat file. @REM tells the computer that whatever follow is only a comment so that it does not try to execute the line. So if you remove the @REM and leave the rest of the line as is, the computer will try to execute the line. So for me, my current webui-user.bat file (after editing it) in the Forge folder looks like (but not exactly, be cautious to type your own drive where A1111 is installed (C: ? D: ? Else ?), and change the with... well... your own username on your computer (no need to tell me) @echo off set PYTHON= set GIT= set VENV_DIR= set COMMANDLINE_ARGS= @REM Uncomment following code to reference an existing A1111 checkout. set A1111_HOME=C:/Users//stable-diffusion-webui @REM @REM set VENV_DIR=%A1111_HOME%/venv set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^ --ckpt-dir %A1111_HOME%/models/Stable-diffusion ^ --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^ --embeddings-dir %A1111_HOME%/embeddings ^ --lora-dir %A1111_HOME%/models/Lora call webui.bat Note : I tried to include (uncomment) the set VENV_DIR=%A1111_HOME%/venv line, but due to some tweak I had done on A1111 to use xformers for speed, it mess up with the Forge which has no xformers installed by default. When I have time I'll try to compare my A1111 speed with Forge: without the A1111 references, with A1111 references... and maybe by adding xformers into Forge

[@cavalrycome]

@@G0odGames Look inside the webui-user.bat file. There are already comments there that explain how to do it.

[@Medychar]

​@@G0odGames you need to uncomment these lines and use your a1111 path: set a1111_home="your_path_to_stable_diffusion_main_folder" set venv_dir=%a1111_home%/venv set commandline_args=%commandline_args% ^ --ckpt-dir %a1111_home%/models/stable-diffusion ^ --hypernetwork-dir %a1111_home%/models/hypernetworks ^ --embeddings-dir %a1111_home%/embeddings ^ --lora-dir %a1111_home%/models/lora

[@Minami1317]

Thank you so much, been using txtimg and transferring all images to SVD for videos is so convenient!

[@KkommA88]

Using it since yesterday, really neat and indeed a bit faster as promised!

[@sebastiankamph]

Nice to hear! What else do you think about it?

[@ClashMafia]

@@sebastiankamph it made me able to run sdxl on my laptop with 4gb vram....need few update as thier are bugs like photomaker not working

[@pete7049]

Thank you for this video. I dont know what was wrong with my A1111 settings, but on a 3090, I am generating like 300% faster! What used to take 4 min 30 seconds now takes less than a minute and a half. Game changer.

[@cherminfoy]

my god, this runs way better than automatic1111, so many errors trying to run automatic1111 and no easy explanation on how to fix them, forge is a life saver thank you!

[@cherminfoy]

ok so do not just start installing extensions, realized when he says xformers and such are not in forge that some extensions may cause issues with forge, killed my forge already but as long as you dont install a bunch of extensions that use xformers (if thats what killed mine i dont know im guessing) you should be fine?

[@kingzamn45]

​@@cherminfoy extensions installed from within forge's extension tab work well for me. Have got photomaker, reactor, sdxl styles and aspect ratio helper

[@AyalSolav]

can you direct forge to the directories of A1111 like you can do with comfy? so you don't need to duplicate every model\lora?

[@Steamrick]

set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@rasyplus]

​@@Steamrickin run.bat right ?can you share full command in run.bat ❤

[@mperspectiva]

The best Video of the year for SD Forge users , outstanding helpful

[@sebastiankamph]

Glad it was helpful!

[@NithinBalakrishnanIsOnline]

Speaking of Knights, What do you call a Knight who is scared of a Chainsaw? A Man with common sense.

[@sebastiankamph]

😁

[@ehrenlos161]

You shouldn't be afraid of the chainsaw, that's everyone else's job.

[@joshuaflair6819]

Thank you, with an RTX 2060 6GB, I was waiting 6 to 8 minutes for a 1024x1024 generation, now it's less than 2 minutes!!! It's day and night !! thank you to the developers of FORGE and thank you for this video

[@sebastiankamph]

Glad it helped!

[@narcizyzzt]

This video came perfectly, I just downloaded it and didn't know how to use it :)

[@CoconutPete]

After seeing some impressive Flux examples on Reddit, I've returned to give Forge another try. I feel sad and nostalgic for SD 1.5 though 😢

[@sonbaz]

Hi thank you for the video. Do i need A1111 installed first? Or do i need to remove A1111 before installing Forge? Or can you have both installed?

[@sebastiankamph]

You can have both. They are standalone

[@TheSslawek89]

The problem is with InstantID. In the preprocessor section you cannot select: instant_id_face_embedding. Because of which the end result is miserable when it comes to this function. I am waiting for some update

[@soullessrights]

Yes, it works I really like this!!! My mistake was checking the download by accident download of zip instead of the "exe" program end down after the download I didn't know where the files were isn't A1111, but the username file is "forge" it worked great I like the new one.

[@CoconutPete]

lora's not showing up... in the A111 version I simply had to delete the cache.json file but no luck with forge.

[@JackPaarthurnax]

The first video where it actually worked! Thanks!

[@sebastiankamph]

I'm happy to hear that! :)

[@PaulRoneClarke]

Installed this both ways and get the same error when it's about to open the WEB UI "Cannot import name 'undefined' from 'pydantic.fields'" Press any key to continue.. Seem to have been a bug reported yesterday as I write this (noticed 2 people reported the same thing on Git yesterday - 17th April)

[@thehmc]

Same error for me. Hopefully someone has some insight on this.

[@defiantfaith324]

Sebastian, I have A1111 already installed should I install the python and git all over again?

[@Alienfodder]

No, you don't need to reinstall them. I installed Forge with the versions of GIT and Python that I already had and it works perfectly.

[@defiantfaith324]

@@Alienfodder Ok thanks bro ..TFI 😄

[@XNazgulXX]

"SVD" is completely missing. how do I get it? Literally can't find how to do this anywhere.

[@DrunkenArrow]

same problem here :/ you found a solution?

[@lennoyl]

Forge is exactly what I was looking for: as easy to use as a1111 but much better optimised. the mask in controlnet is the thing I was looking for for a long time in a1111 (that and a better handling of the VRAM) and why I spent more time on comfyUI. There are a lot of features "integrated" I don't even know what they do ^^

[@sebastiankamph]

Time to start exploring then! :D

[@souravroy8834]

@@sebastiankamph hi can you make a video about installing in amd gpu because i have tried it shows me error

[@mytwocrazydogs9715]

Is there a way to have Forge use the extensions from Automatic1111? or do I need to just copy the contents of the extension folder over to Forge OR just reinstall all extensions in Forge UI? I see all the other ways to link/reuse items from Automatic1111, but not seeing anyone ask about the extensions. Thanks!

[@kingzamn45]

Runs flawlessly on my gtx 1060 6GB version and compared to EasyDiffusion, which is supposed to be the fastest, this is so much more faster than that too and so much more feature rich. A full 720p image upscaled to 4k after HiRes fix added as well takes like only 45 seconds to a minute. If you use his LCM tutorial with the 1000 times faster Lora, then that time is cut in half and you can even directly renders high resolution images even quite quickly. Also the 1 click installer worked fine for me.

[@prixmalcollects9332]

the problem is when you use sdxl model and use lora, the generation is very slow, does it make loras with sdxl faster? I have 4090, etc but generation is very slow with loras

[@ronaldbruneau1980]

I've had no luck with the speed up lcm loras using the 4090 with multiple models.....

[@prixmalcollects9332]

@@ronaldbruneau1980 so it remains slow? Guess we'll have to wait for updates

[@ronaldbruneau1980]

I was actually referring to a1111 deforum using the loras in animation. They actually slowed me down with the 4090

[@Elwaves2925]

What does Sebastian call going on holiday with his family? Kamphing. Started using this earlier in the week and it's great. Not perfect yet but great and worth it for the performance increase alone. The two things it needs for me to swap full time, a fix for ControlNet not detecting IPAdaptor and for CivitAI Helper to be updated to work with the tree view Forge uses for loras etc.

[@sebastiankamph]

Tell me more about this ipadapter bug

[@Elwaves2925]

@@sebastiankamph It's not picking up the preprocessor file, the one used for face swapping, so you can't select it. The github has several people mentioning this and it's labelled as high priority. I think it may have been fixed too (not sure) but not made it to the main release yet. 🙂

[@CoconutPete]

Ok, I'm back and have come to the conclusion for me (a SD noob) I'm going back to A1111 simply because there's not enough documentation for bugs involving getting various extensions/models to work properly with Forge. But, this is probably a good opportunity for your channel to make Forge content, because there's not much out there. I may try again in the future. Also I found a post on Reddit which apparently there is some bickering between ComfyUI developer and Forge claiming that one stole code from comfy.. not sure all the details, but might be worth a look lol

[@bigboss3279]

bro thanks to you i have 2-3 times faster generation on sdxl with my rtx 4050, the guy who made forge is genius

[@RisetOnline]

Hello friend, have you updated your Forge to the latest version? Any issues with the extensions? I'm too scared to update Forge because the latest version is causing a lot of errors, and the author announced that they won't be updating Forge anymore. They suggested going back to A1111 🥲

[@johnny-mnemonic13]

@@RisetOnline oh wow, so not a good idea to install forge anymore?

[@fernando749845]

Wonderful Sebastian, thank you! Do you know why many ControlNet models for SD1.5 still haven't been created for SDXL? For example the Tile model. I also feel that models like Depth don't work as accurate or create strange results.

[@sebastiankamph]

You're right that the 1.5 controlnets give better results. Not sure why the XL ones are so limited.

[@mlsterlous]

Automatic111 works great in directml mode for amd, and i don't even have GPU, only ryzen 7735 integrated graphics. Not only it runs 1.5 models, but also SDXL. 512x512 in 50sec. Something like 1200x800 4-5~ min. I think it's great speed.

[@nenadm5747]

What do I have to do to make it work on Amd card?

[@danishraza8843]

explain with details please

[@FunWaku]

@Sebastian ... but why the forge folder size keeps getting bigger and bigger everytime we prompt or generate something ? any ideas ?

[@esperlihn4446]

I assume probably because it's saving all the generated images to that folder

[@FunWaku]

same thought at first untill I recgnized that it wasn't that, it's the Nvidia library files and model@@esperlihn4446

[@oOROcKetguYOo]

How to get it to work with AMD GPU ??? At the end of the Advanced Install (and after adding --skip-torch-cuda-test) there is only the information "Please check that you have an NVIDIA GPU" ...

[@parryhotter3138]

same

[@snaileri]

@sebastiankamph How do I add upscaler models, like 4x-UltraSharp to Forge? adding it them to .../models/ESRGAN/ doesn't work. The upscaler doesn't appear in the SD Upscale -script list.

[@dm204375]

Forge is amazing and a lot faster then base Auto1111, but there's a caveat. You can't use hypernetworks and some extensions and it has plenty of its own bugs so its best to install it in tandem with regular Automatic1111.

[@Phraxas52]

Which extensions can't be used? There seems to be a virus going around the world preventing people from explicitly stating what extensions do or don't work.

[@CoconutPete]

sorry for questions but anyone know if you can use SSD-1B with forge?

[@Avalon19511]

Quick question since I'm not a patreon I installed Forge manually and I noticed there is no update.bat, so what do I do in that case?

[@musicinthemachine]

I installed it a couple of days ago, but ControlNet doesn't work. Any time I try to use it, no matter which type, Canny, Lineart, etc., it gets an error saying: TypeError: 'NoneType' object is not iterable. I put the models in the folder and they show up in the drop down in ControlNet, but always get this same error.

[@sebastiankamph]

And you put an image in as an input?

[@musicinthemachine]

@@sebastiankamph Yes. I have both A111 and Forge installed and set up the same way. It works in A111 but not in Forge. Everything else in Forge is great. It is definitely noticeably fast even on my laptop with 16 GB VRAM. I would love to get the ControlNet working as well. I am not a total expert, but I am the creator of the Acorn Is Spinning models on CivitAI, so I have quite a bit of experience using A111 and some with ComfyUI.

[@Steamrick]

Try referencing your controlnet folder in the A1111 install instead? set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@fretts8888]

Not at my machine rn, but I think control net defaults to models prefixed with “lora”, not sure what they are but they seem to generate this error, select one of the other models to see if it works (does for me).

[@musicinthemachine]

@@sebastiankamph I found the problem. I thought that I had all of the SDXL ControlNet models in the folder but it seems somehow that only the 1.5 models were in there. I was using my Acorn is Spinning XL V2 model and it was trying to use the 1.5 ControlNet models. Now I have it working. Overall I am loving Forge.

[@stableArtAI]

Not sure why looking at install it show Total VRAM 65536 MB, total RAM 65536 MB, yet have 16GB VRAM on a 64GB system

[@MonkeyDLuffy-dx6ow]

Thanks man great video for beginners 🙏🏼

[@Liqwalker123]

ok i see the pros of forge but what are the cons? do i miss any options out in forge then using automatic1111? i want to make images like a comic with faceswap of real photos and keep the face, clothes and style but want to be able to change the pose and the background

[@jgillman1984]

When launching the webui-user.bat file I get: File "F:\Ai\StableDiffusion\Forge\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\params.py", line 4, in from pydantic.fields import FieldInfo, Undefined ImportError: cannot import name 'Undefined' from 'pydantic.fields' (F:\Ai\StableDiffusion\Forge\stable-diffusion-webui-forge\venv\lib\site-packages\pydantic\fields.py) Any suggestions?

[@thehmc]

Same error. Just commenting to get notified if anyone solves this.

[@-Belshazzar-]

but how can one install Forge over an existing 1111 install, so it will use the already hundreds of GB of models that are already in my hard drive?

[@Asaghon]

Using "run" after "update" didn't work for me, gave me an error. I had to go into the webui folder and install requirements first with cmd. Then it worked

[@evelynstrip]

uf.. I just installed A1111 in terminal.. with all the cmake protobuf rust python git wget stuff.. now I have no idea how to uninstall all of this.

[@benharris144]

Hey, you did it!

[@sebastiankamph]

I got you! What do you think of Forge?

[@benharris144]

@@sebastiankamph Loving it so far! When generating tons of images, that extra 5-9 seconds per image REALLY helps.

[@Elwaves2925]

@@benharris144 Agreed, when using wildcards those few seconds really mount up and it can save a LOT of time.

[@mlsterlous]

NEED tutorial for AMD in directml mode. Actually i almost got it running, but... when i press generate, there is an error, type error or something....

[@arch4466]

@user-ef2rv9el9x yo pc js sucks lil bro

[@mr-peabody]

Hi Sebastian How big is the "one -click package" and can I run it from a thumb drive? Thnx

[@TheCynicalNihilist]

Thanks, ill be using this to install forge later today

[@sebastiankamph]

Let me know what you think of it!

[@TheCynicalNihilist]

@@sebastiankamph works really well, I just got a 4090 today and its a lot better structured, faster, ill have to play around with it. BUT I wanted to ask you, I tried adding your styles prompts but I cant find where in the webui folder to put it? theres no styles folder ive looked everywhere.

[@johnlonggone]

​@@TheCynicalNihilistthere is not specific folder for styles. Just drop it in the general stable diffusion webui folder.

[@malarkey102]

Hi I installed forge but don’t know what all the extensions are it came with I don’t know most of it

[@seanknowles7987]

can you copy your fooocus model /checkpoints/lora folders etc from fooocus folders and port them over to your forge folder? Or do I need redownload them for forge specifically? Also, is this download access different than forges webUI ? Or is accessing the webui for forge a simpler method?

[@BenvanGeest]

Thanks so much for these guides, Sebastian. Automatic 1111 worked great for me, but somehow now I get the following error when trying to generate images: TypeError: 'NoneType' object is not iterable. Do you have any idea what this error message refers to? Cheers, and keep up the great work!

[@ronaldbruneau1980]

I use forges deforum tab to create animations. I would like to know how to create the animations within a boundary. I projection map so I would like to know how to keep the animations within the map of my house. Would you know how to accomplish this? I have a png map file that I created but unsure what to use it with. TIA

[@SmartNeto]

Since your vladui video I still using that, it's a lot faster than a1111 (at least for my AMD processor), but honestly it's a bit unstable with another extensions. I have an RTX 3070 and a Ryzen 7 5800X, would forge be better for me?

[@sebastiankamph]

All signs point to yes. But your mileage may vary.

[@SmartNeto]

After a bit of testing I can say that this was actually a great upgrade for me, thanks for your amazing content (:

[@sebastiankamph]

@@SmartNeto happy to hear it!

[@thays182]

How do you properly direct forge to use your SD1.5 Folders for Models/Loras/etc?

[@MrDebranjandutta]

How do I share models between forge and the standard install of automatic 1111?

[@yyzjuan99]

Does this intallation works without GPU and just CPU?

[@Bodrick_G]

Can you forward dir to A111's old files (checkpoints, loras etc) to save space?

[@sebastiankamph]

Yes! It's in the webui-user.bat file

[@Steamrick]

set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@bluemurloc5896]

do we have to insert the command git pull into the .bat to make sure it updates automatically?

[@aritradey8270]

I have 2gb of VRam will this work on my PC?

[@NewPersonBorn]

You need 4gb

[@darkman237]

Why would it ask for a username? Never had that happen before...

[@angelalmalaq]

i have bizard question ? do forge will detect my old stable diffusion and integrate it self, or it create a new stable diffusion !!!!

[@jettro8523]

IF you have setup A1111/Comfy/Invoke then its probably safe to use the one click method, first time a one click worked flawless! Great video Seb!

[@mistercapitale]

Hey, any chance we can see what all the tabs do? Examples maybe?

[@classichungamatv7012]

what are the requirements specs to use a1111

[@sebastiankamph]

Nvidia gpu 4gb+ vram

[@classichungamatv7012]

@@sebastiankamph i have nvdia gtx geforce 1650 with 4gb dedicated gpu memory, can i run it?

[@classichungamatv7012]

@@sebastiankamph ?/

[@classichungamatv7012]

@@sebastiankamph ??

[@johnquertinmont9215]

Can I save all my checkpoints, LORA's, wildcards, and textural inversions to a external drive, then put them back after I install Forge, because I have so many of them and I don't want to lose them

[@copester1204]

Great video but you forgot to include how to configure it to use the models, loras, etc., we've downloaded for A1111, ComfyUI, Fooocus, etc.

[@sebastiankamph]

You're right! It's available in the webui-user.bat

[@CoconutPete]

​@@sebastiankamphcan you explain more about this bat file if you have time someday?

[@Steamrick]

@@CoconutPete set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@berke9313]

I dont know if you will answer, but I read that you can upgrade your A1111 to FORGE A1111. Can you make a video for that, pls

[@youzzef.]

Hello anyone know how to stop stable diffusion from using hdd i installed it on ssd but it using hdd seeing it on task manager

[@icchansan]

Thx for sharing this tut, where can i download the stable video checkpoint? try to find the one u used and no luck

[@sebastiankamph]

Check one of my latest videos and it should be there

[@Alpenmilch]

Forge is indeed faster but I have the ongoing issue that my chekpoints, preset, lora, ect have a hard time to load. I have to refresh every single time when I boot it. It seems like the interface is loaded before that the extensions\models are loaded in.

[@sebastiankamph]

I'm sorry to hear that. Not sure what could be the issue.

[@ВсякоеРазное-д6я]

Thanks for the video. Can I ask you the dumbest question? When I install it, I always get a light colored theme. All the videos have a dark interface. Is it some theme that needs to be downloaded or browser installation? Thanks.

[@sebastiankamph]

It's based on your browser settings

[@soulife8383]

I'm just getting started with my journey into Stable Diffusion. 1 thing that isn't super obvious to me is, I understand that Forge, A1111, etc, are just browser front ends for SD, but when you follow these install guides, does it bundle with SD? Or am I supposed to install SD as per their website and then follow these guides? Edit: and also, is it safe to have multiple front ends installed?

[@sebastiankamph]

You can install multiple front ends. SD is just that .safetensors file the front end reads. So you're good with just installing a1111, Forge, Comfy etc.

[@부정선거4.15]

Im using sd 1.5 . I wonder . Can i use the lora for 1.5 version in the forge?

[@mohdfaredzuanmohdnoor4008]

I have questions. If I already installed A1111, do I need to install python & git again for forge? And can a pc have both forge and A1111?

[@Elwaves2925]

You don't need to install them again. If you use the 1-Click installer it will do it all for you and everything will be in a self contained folder. So yes, you can use this alongside A1111, it's what I'm doing until a couple of Forge things get fixed.

[@duskairable]

Forge wont install them again if you have already installed. I install forge manually with git clone btw. You can have both.

[@FerdinandvanDam]

I do not get it running on my Mac studion M2ultra

[@FunWaku]

If anyone reads this , DO NOT TRY IT UNTILL YOU KNOW HOW IT WORKS .... To me it works perfectly fine but for those who own 16GB of ram , leave to LIVE !

[@tomasdvoracek6098]

why?

[@FunWaku]

@@tomasdvoracek6098 it eats your PC resources as if it has endless void hunger, well atleast from my experience several weeks ago !

[@DamielBE]

Can it use existing checkpoints and lora folders with Automatic 1111? I already have a crapload of these, and I don't want to copy all them to another program's folder tree

[@sebastiankamph]

Yep! I explain how in the patreon guide. It's fairly simple, check the webui-user.bat file.

[@augusttinahell]

Thanks a lot, great info!

[@sebastiankamph]

Glad it was helpful!

[@lxic-bz8hf]

anyone know how much space does it take on hard drive ?

[@CoconutPete]

I assume you need to make a brand new installation? I already installed A1111 ... dang 😢

[@sebastiankamph]

Yes and no. There's instructions on github on how to use the Forge fork from a1111. But I would honestly just install this again and have both.

[@CoconutPete]

@@sebastiankamph I'm giving it a try and installing now... I assume I can copy loras and other stuff from a1111 into forge installation... will report back as a potato pc user lol

[@Steamrick]

@@CoconutPete Just reference your existing a1111 install in the webui-user.bat set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@VladPopaYT]

Do you know how to add instructp2p in my controlnet?

[@Shogox]

You can also direct it to the model folder used by a1111.

[@sebastiankamph]

Great tip!

[@0ZeroTheHero]

How do you do that, is there a tutorial online you can reference to?

[@Steamrick]

@@0ZeroTheHero set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@RegularAdventurer711]

I haven't seen the fooocus video completely but is it possible to have both forge and foocus ?

[@Liqwalker123]

why not? you can have fooocus, forge, automatic11111 and comfyui if you want to

[@tvandang3234]

How can I reference the existing model folders that are in my A1111 for Forge without making another copy?

[@sebastiankamph]

This is available in the webui-user.bat file. There's a preset already for it. Just remove the @ and add your path to a1111

[@Steamrick]

To reference everything at once: set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@rodrigobarrosempreendedor]

Great video. Can you use sd3 webui forge?

[@Steamrick]

As a fellow 4090 owner I don't care about the speedup, but a 1.6x max resolution increase is more than enough reason to switch over. I could barely do 3440x1440 while maxing out vram and system ram before and using system ram caused a huge slowdown. Maybe now it'll be much faster and even enable 4K resolution? I gotta try it out later.

[@sebastiankamph]

Agreed! For still image generations, 4090 is fast enough. The reason I even bought it was for the extra vram.

[@ronaldbruneau1980]

Are you doing animations? If so please post if you are getting better results. I'm running the 4090 and I'm undecided if I want to install this yet. I'm making text animations about 1min long approx 40ish min in deforum 1920x1080

[@Steamrick]

@@ronaldbruneau1980 I've only done some very few experiements with a Comfy workflow, but animations / video is too clunky for me for now. Now if there were an open source model that has results like Sora...

[@hotlineoperator]

Be aware, because I will try everything you advise.

[@sebastiankamph]

Then I'll be sure to advise lots of extra stuff just for you! 😅

[@steel146]

I have successfully installed Forge. However, the controlnet models don’t show up. I have experience in A1111 and installed model files in the proper folders. Does anyone have same problems?

[@Steamrick]

Try referencing your A1111 folders: set COMMANDLINE_ARGS= --forge-ref-a1111-home "[insert-path]\stable-diffusion-webui"

[@Tomasin19]

Lemme guess... doesn't work for AMD, right?

[@lennoyl]

it is possible to run a1111 with AMD GPU, so I guess it is possible too. Look at a tutorial on youtube for using AMD GPU with a1111 and adapt it for Forge Diffusion.

[@mlsterlous]

​@@lennoyli have no problem running it with automatic111. But it's not so obvious with forge. So we need a tutorial.

[@user-he3gn5bb2y]

try adding: --directml --skip-torch-cuda-test to your webui-user bat file, letus know how many iterations/sec if it works

[@mlsterlous]

@@user-he3gn5bb2y i tried it already of course. It launches without any errors, but after pressing "generate" there is a typeerror or something.

[@sub-jec-tiv]

Little confused. If you already have a Python install from running A111, do you still need to reinstall Python if you do manual install? Or can you just pull the forge data and start there, using the same Python setup you already have?

[@ronaldbruneau1980]

Read through the comments but short answer is no

[@godspeed2145]

Python that works for a1111 is the same exact one

[@kevinramos3479]

inside forge does xformers matter?

[@onyxfox.09]

I dont have the SVD tab

[@DrunkenArrow]

me neither, found a solution?

[@onyxfox.09]

@@DrunkenArrow sadly no. It turns out that it’s not supported in the newer versions.

[@DrunkenArrow]

@@onyxfox.09 jeah, thats what was on my mind.. ill try to get an older version later or find another option :D ty for your answer

[@brunodegiovanni6428]

Fantastic with my 3060 6Gb! Anyone found a way to fine tune ControlNet weight in the XYZ script? With A1111, the [ControlNet] weight appears amont the options in the drop down menu, but with Forge doesn't

[@edwardferry8247]

If you need a cat in a hat in a hurry this looks good 👍

[@brandonavelino]

cand forge run on AMD GPUs? any config?

[@joem2042]

Is Forge more stable than automatic1111?

[@daadirsjr5755]

iam runing into black pictures please any help