You have explained it the best way than anyone else... This is for the first time I understood all the basics as to how everything actually works, why you download the models, why you put every cpkt files where... everything is clear now! A big thanks!
Your videos are amazing. I went from no idea what it even was last monday to now trying to train an embed in less than two weeks and you are a big reason for it. Thank you!
Laura can you please do a video about using codeformer to improve facial quality in a local Mac environment? I am a great fan of your videos and blogs and very thankful for your inspiration. Please keep putting these videos out!
Laura, thank you so much for your videos! They've been so helpful. I do have a question regarding downloading models. I'm trying to download a model from CivitAI but the only type of file that is available for some models is a SafeTensor filetype. I don't see any with the ckpt extension. Will it work if I put the SafeTensor file in the /models/Stable-diffusion folder?
I really loves, it's eye opening. I run an ecommerce website, do I will be using it to generate product mockup for my store. Can you please make a video on how to setup stable diffusion on deepnote. I just discovered that deepnote is free, but I'm not a programmer
Hi, yes, lots are having the same issue. Some upgraded to Colab pro and don't get the error anymore. If you don't want to upgrade, you can try running the below code before the requirements cell: !pip install pyngrok
This is great!! Thank you so much! When I try to upload a lora safetensor into the lora folder, It won't let me, and I get an error message back saying "unexpected token". Does anyone know why this is?
Hello, when I take the download link of some models like Juggernaut XL from civitai and put the link in the MODEL_LINK, I get an error for downloading(HTTPError: HTTP Error 401: Unauthorized ), how can I fix it? I'm sorry if I didn't convey my meaning well, I'm not an English speaker :)
Nice video 💯. Can you please tell me what is the difference between running sd on Google colab and running it locally? Is it both same performance wise?
It depends on what computer you are using. If you have a mac, usually Colab/Runpod is better. If you computer is windows (Nvidia), then you'll probably want to use it!
I'm facing an issue regarding the connection. The colab disconnects only after few minutes of connecting and before I click on generate or in the middle of the process it disconnects showing "Runtime disconnected" and I'm unable to use it. It was working perfectly fine but from 2 days I'm facing this issue. Please let me know how to stay connected??
Hi, Colab is not fully reliable with connection, I understand (mainly if your internet connection is not stable). If the issue is your internet connection, you can install SD on your computer locally. Otherwise, you can swap to RunPod, or Paperspace and see how it goes. I think Paperspace has a free option, but I have never used it.
So, if I understand correctly, anytime Colab disconnects, you gotta do the whole process again, right? It is very time consuming for me (I only use a tablet so I rely mostly on renting GPUs). Run Pod and Paperspace operate like google colab if I understand you correctly?
hey, question.... I have an M1, and would like to keep a bunch of models and checkpoints on an external drive. Can I just make a shortcut in the models/Stable-diffusion folder that points to the files on my external drive?
Hi, I have never tried tbh. I think you can do that, but you might need to have the all stable diffusion webui folder in the drive. This is because the .ckpt or safetensors have to be inside the models folder.
My question is How we can add 3 train people in Single CKPT like if I mention in Prompt Alice, Jane, Rachel taking group photo now in Single train model we see twins instead of seperate individuals. now with Celebrities but not with out custom models. So is there any technique to compile all our Train models in 1 CKPT?
You can train more people in one single ckpt; just difficult to check the results and adjust parameters if you need to train the model again. This is because the model might learn one person differently from the other. So, I would suggest to train one model for each person
Hi. I'm having problem with ControlNet part. It looks a little bit different cause I think it was updated(it has two bars now). And no matter what I click it shows me some error and I can't continue. Can you help please?
Thanks for sharing this tutorial. I have encountered a problem and I am wondering what I have done wrong. I have trained a traditional stable diffusion + dreambooth model (i.e. not using this faster version) via Colab by using photos of me at a google account 1 and let's call it model 1. Then, I use another account, account2, to install a SD+auto1111 via Colab as your previous tutorial showed. Then, following the steps of this tutorial, I loaded the model 1 to the auto1111 host by account 2 and generated images. The images are way distorted and don't look like me. I have to go back to my account 1 to run model 1 to get it work. Could you explain to me what I did wrong ? Couldn't we load ckpt model at another account ? I thought the auto1111 is just a interface to run models we selected. Thanks
Hi, as you said, auto1111 is just an interface, so you can run any model. Have you run the same model on both account1 and account2 and got different results? If yes, then you are probably using different settings in the model (e.g., seeds). If no, then it can be that you need to retrain your model using different learning rate/training steps. Also, remember that when you build the prompt you need to use the filename of the images you used to train the model (i.e., if you trained the model with files named "laura (1).png", "laura (2).png", "laura (3).png", etc., the prompt should be "a photo of laura....
When I am about to run "AUTOMATIC1111 repo", this warning appears "You may be executing code that is disallowed, and this may restrict your ability to use Colab in the future. Please note the prohibited actions specified in our FAQ." what happened? should I continue the process or cancel it?
I have some issue. I already tried to install 3 TIMES. I already deleted the sd folder on g-drive and reinstalled. I just pressed play on each cube. I first I getted the error: No module named 'pyngrok" That I tried to solve putting the line: !pip install pyngrok on the first line of last block But still have the error : can't open file '/content/gdrive/MyDrive/sd/stable-diffusion-webui/webui.py': [ The error occurs on the "Start stable diffusion" block Please help...
@@LaCarnevali Thanks for the feedback, No, I have not a pro account, , but since the full error is: can't open file '/content/gdrive//sd/stable-diffusion-/webui.py': [Errno 2] No such file or directory maybe can be something relative to the installation or execution script?
Hi monster, if you installed different models previously, and they are stored in the Drive, they should appear automatically when initialising stable diffusion.
webUI is ban on colab, they recently updated their policies. This code won't work anymore. I just tried and figured it out. I guess bad luck for all of ous.
I can still run in Colab. Anyway, if you have issues, you can try RunPod or Paperspace. I think the latter has a free option too (but not 100% sure as I have never tried it)
@@bondzh I think you can try paperspace, it should have a free version, but not 100% sure as I haven't tried it yet. Otherwise, you can download it locally.
@@LaCarnevali Thanks. I tried to download it locally, but how can I make my Intel Graphics 630 GPU work on Pytorch? Very surprisely, the GPU works on cloud version but cannot work on local version.
weird how using AI to perform tasks involves a system of this many steps and manually typing things into a non GUI interface like I am on my amiga in 1992 installing things manually...
youtube deleting my comments why? you don't have to upload model to Gdrive, you can put direct link(copy link after you inititate download of file from civitAI or huggingface and then canceling it) in the download model section, works for LORA also, but you need to move LORA model to LORA folder since it will be downloaded in the model folder, you can also just rename the downloaded file in Gdrive since it will probbably be named wrong.
i have bad internet so i figured out a way to bypass the uploading/downloading part while following your first guide, i'm on my first try right now :D, this comment is to expand my ego to new heights as i compliment my own intelligence xD.
note that if you download LORA the code in the downloader will say wrong file or link or whatever, but that's nonsense, it still downloads it to Gdrive, so just move it, rename and use it.
Hi Ad Mortem (interesting nickname by the way lol), yeah you can, but if you want to upload them all in one go, you cannot do that, unless you run the model twice, or three times, etc.
Hello. I uploaded three models into the google drive folder, did the exact steps and got this specific error. What should I do in this case? NameError Traceback (most recent call last) in 12 13 if Redownload_the_original_model: ---> 14 with capture.capture_output() as cap: 15 rmv() 16 NameError: name 'capture' is not defined
Hi! It seems that "capture" was not imported. Are you 100% sure you run the "Install/Update AUTOMATIC1111 repo" section? That includes the below code, which is needed to run capture: >>> from IPython.utils import capture
You have explained it the best way than anyone else... This is for the first time I understood all the basics as to how everything actually works, why you download the models, why you put every cpkt files where... everything is clear now! A big thanks!
Your videos are amazing. I went from no idea what it even was last monday to now trying to train an embed in less than two weeks and you are a big reason for it. Thank you!
Super nice! Thank you! :)
Hi Laura, I would like to thank you for providing such great tutorials. Keep up with the good work.
thank youuuuu i appreciate the straightforward tutorial it helped a lot!
Thanks for your enthusiasm and enjoyable videos
Can’t get mine to work… 🤦♂️ Maybe mine isn’t powerful enough. Mac Mini - late 2012 - 16g RAM - Ventura with OpenCore.
thanks for making a video tutorail on the qustion that i asked thnks again
Laura can you please do a video about using codeformer to improve facial quality in a local Mac environment? I am a great fan of your videos and blogs and very thankful for your inspiration. Please keep putting these videos out!
Thank you, Hans! Sure, will add it to my to-do list ;)
Great!
Was very helpful, thanks.
I am sure you get asked all the time but are you planning on doing a video teaching how to train models?
Yes! I am going to… very soon :)
@@LaCarnevali Thank you! I'll be waiting for it!
great video as always, thanks :)
Thank you, James 🙏🏻
Laura, thank you so much for your videos! They've been so helpful.
I do have a question regarding downloading models. I'm trying to download a model from CivitAI but the only type of file that is available for some models is a SafeTensor filetype. I don't see any with the ckpt extension. Will it work if I put the SafeTensor file in the /models/Stable-diffusion folder?
Yes, safetensor and checkpoint are the same thing. Safetensor files are safer, cannot contain malware.
So helpful, Laura! thanks!
Hi Laura, thanks so much for the video! Is it possible to download a refiner? This would be so helpful, please please!
I really loves, it's eye opening.
I run an ecommerce website, do I will be using it to generate product mockup for my store.
Can you please make a video on how to setup stable diffusion on deepnote.
I just discovered that deepnote is free, but I'm not a programmer
Ho notato il tuo accento italiano e dopo aver visto il tuo nome ho avuto la conferma. Fa piacere vedere altri in Italia interessarsi a queste cose :)
Thanks , that really helped me !! :D
Very helpful, I love your accent too!
Hi! Where is the link of Google Colab Notebook? I love your videos!
colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb
Oops, sorry - I ll add it now. Thanks for spotting it! :)
@@LaCarnevali❤
Thanks for the great tutorial and video,
what about installing lora and other extension on colab?
How to fix this error
ModuleNotFoundError: No module named 'pyngrok'
Hi, yes, lots are having the same issue. Some upgraded to Colab pro and don't get the error anymore.
If you don't want to upgrade, you can try running the below code before the requirements cell:
!pip install pyngrok
@@LaCarnevali not working 😞
You need to upgrade :(
There is no longer a checkbox called re download model, what should I do?
Very helpful thank you
This is great!! Thank you so much! When I try to upload a lora safetensor into the lora folder, It won't let me, and I get an error message back saying "unexpected token". Does anyone know why this is?
Hello, when I take the download link of some models like Juggernaut XL from civitai and put the link in the MODEL_LINK, I get an error for downloading(HTTPError: HTTP Error 401: Unauthorized
), how can I fix it?
I'm sorry if I didn't convey my meaning well, I'm not an English speaker :)
hey great video but when i try to switch models, it reverts back to the one i had selected before and im not sure why
What model is that?
Hello Laura thx for video, in Start Stable-Diffusion tab i have "Use_Cloudflare_Tunnel" no "Use_localtunnel" as in your video. Why?
Hi Cristian, the notebook is constantly updated. Cloudtunnel still give you a secure no public connection :)
Just now i check there is no Redownload original model checkbox
how did you find the solution
@@mertolojis no solution, i end up installing it locally on my pc
I found the solution, just add it as path@@HeinleinShinobu
Nice video 💯. Can you please tell me what is the difference between running sd on Google colab and running it locally? Is it both same performance wise?
It depends on what computer you are using. If you have a mac, usually Colab/Runpod is better. If you computer is windows (Nvidia), then you'll probably want to use it!
I'm facing an issue regarding the connection. The colab disconnects only after few minutes of connecting and before I click on generate or in the middle of the process it disconnects showing "Runtime disconnected" and I'm unable to use it. It was working perfectly fine but from 2 days I'm facing this issue. Please let me know how to stay connected??
Hi, Colab is not fully reliable with connection, I understand (mainly if your internet connection is not stable). If the issue is your internet connection, you can install SD on your computer locally. Otherwise, you can swap to RunPod, or Paperspace and see how it goes. I think Paperspace has a free option, but I have never used it.
So, if I understand correctly, anytime Colab disconnects, you gotta do the whole process again, right? It is very time consuming for me (I only use a tablet so I rely mostly on renting GPUs). Run Pod and Paperspace operate like google colab if I understand you correctly?
hey, question.... I have an M1, and would like to keep a bunch of models and checkpoints on an external drive. Can I just make a shortcut in the models/Stable-diffusion folder that points to the files on my external drive?
Hi, I have never tried tbh. I think you can do that, but you might need to have the all stable diffusion webui folder in the drive. This is because the .ckpt or safetensors have to be inside the models folder.
Thank you!!
When writing the prompt I don't need to invoke the keyword to generate the image(the trained keyword ), on the computer yes but in google colab no?
should always use the trigger word for loras
Thank you very much and I love your tutorials. I hope to continue learning much more with your channel, Miss Laura@@LaCarnevali
My question is How we can add 3 train people in Single CKPT
like if I mention in Prompt Alice, Jane, Rachel taking group photo now in Single train model we see twins instead of seperate individuals.
now with Celebrities but not with out custom models.
So is there any technique to compile all our Train models in 1 CKPT?
You can train more people in one single ckpt; just difficult to check the results and adjust parameters if you need to train the model again. This is because the model might learn one person differently from the other. So, I would suggest to train one model for each person
Hi. I'm having problem with ControlNet part. It looks a little bit different cause I think it was updated(it has two bars now). And no matter what I click it shows me some error and I can't continue. Can you help please?
What error?
very usefull ! thx
my model is not ckpt .. why ?? my safetensors
its still work ??
yep. checkpoint and safetensors are the same thing. Difference: safetensors cannot contain malware
Thanks for sharing this tutorial. I have encountered a problem and I am wondering what I have done wrong. I have trained a traditional stable diffusion + dreambooth model (i.e. not using this faster version) via Colab by using photos of me at a google account 1 and let's call it model 1. Then, I use another account, account2, to install a SD+auto1111 via Colab as your previous tutorial showed. Then, following the steps of this tutorial, I loaded the model 1 to the auto1111 host by account 2 and generated images. The images are way distorted and don't look like me. I have to go back to my account 1 to run model 1 to get it work. Could you explain to me what I did wrong ? Couldn't we load ckpt model at another account ? I thought the auto1111 is just a interface to run models we selected. Thanks
Hi, as you said, auto1111 is just an interface, so you can run any model. Have you run the same model on both account1 and account2 and got different results? If yes, then you are probably using different settings in the model (e.g., seeds). If no, then it can be that you need to retrain your model using different learning rate/training steps. Also, remember that when you build the prompt you need to use the filename of the images you used to train the model (i.e., if you trained the model with files named "laura (1).png", "laura (2).png", "laura (3).png", etc., the prompt should be "a photo of laura....
should it be in /models/StableDiffusion? not /models/Lora?
might have changes with time
When I am about to run "AUTOMATIC1111 repo", this warning appears "You may be executing code that is disallowed, and this may restrict your ability to use Colab in the future. Please note the prohibited actions specified in our FAQ."
what happened? should I continue the process or cancel it?
Are you running it from a remote desktop or SSH?
is it any problem that .json file on my drive? ,I have accidently put it wrong drive, pls help me to remove it.pls
Sorry, I am not sure I got the issue?
Thanks, lady
thx a lot
helpful, thx
After few minutes is says connection errored out. Any idea on why this happens?
Probably need to upgrade to Colab pro?
@@LaCarnevali I tried, but sadly colab pro isn't yet available for the country I live.
I have some issue.
I already tried to install 3 TIMES.
I already deleted the sd folder on g-drive and reinstalled.
I just pressed play on each cube.
I first I getted the error:
No module named 'pyngrok"
That I tried to solve putting the line:
!pip install pyngrok
on the first line of last block
But still have the error :
can't open file '/content/gdrive/MyDrive/sd/stable-diffusion-webui/webui.py': [
The error occurs on the "Start stable diffusion" block
Please help...
Google banned the free version of colab for running SD. so that might be the issue ... do you have Colab pro?
@@LaCarnevali Thanks for the feedback, No, I have not a pro account, , but since the full error is:
can't open file '/content/gdrive//sd/stable-diffusion-/webui.py': [Errno 2] No such file or directory
maybe can be something relative to the installation or execution script?
@@ikarosound2504 cannot run anything without the Pro. Try ComfyUI, which is free
What If I want more than one models?
Hi monster, if you installed different models previously, and they are stored in the Drive, they should appear automatically when initialising stable diffusion.
Is there a video for LoRA's?
Nope, but will do that soon
it was useful!
i can't add the one i downloaded from civitai
Add it manually in your drive
Nice
pleasee dont u have guide of how to install sd locally on windows?
great videos,keep going
p.s so cute :)))
Very easy on windows, only need to follow this:
github.com/AUTOMATIC1111/stable-diffusion-webui#automatic-installation-on-windows
webUI is ban on colab, they recently updated their policies. This code won't work anymore. I just tried and figured it out. I guess bad luck for all of ous.
need to upgrade to Colab Pro
@@LaCarnevali are you sure it is working for Pro?
Its not allowed in colab anymore. Whats the solution?
I can still run in Colab. Anyway, if you have issues, you can try RunPod or Paperspace. I think the latter has a free option too (but not 100% sure as I have never tried it)
@@LaCarnevali Sorry, the free version cannot run stable diffussion in Colab anymore. Paid version still okay. Any solution?
@@bondzh I think you can try paperspace, it should have a free version, but not 100% sure as I haven't tried it yet. Otherwise, you can download it locally.
@@LaCarnevali Thanks. I tried to download it locally, but how can I make my Intel Graphics 630 GPU work on Pytorch? Very surprisely, the GPU works on cloud version but cannot work on local version.
is it possible to add Lora ?
Yes, in the Lora folder
weird how using AI to perform tasks involves a system of this many steps and manually typing things into a non GUI interface like I am on my amiga in 1992 installing things manually...
not working now ı guess
I AM UNABLE TO INSTALL STABLE DIFFUSION ON COLLAB., ANYONE HAVE SAME ISSUE?
Google banned the free version of colab for running SD. so that might be the issue ... do you have Colab pro?
@@LaCarnevali thanks so much for the reply, no, I have not the pro version, so I cannot use it for free?(i have potato pc)
@@ikarosound2504 nope, Colab Pro only. You can try ComfyUI, which is still free
youtube deleting my comments why? you don't have to upload model to Gdrive, you can put direct link(copy link after you inititate download of file from civitAI or huggingface and then canceling it) in the download model section, works for LORA also, but you need to move LORA model to LORA folder since it will be downloaded in the model folder, you can also just rename the downloaded file in Gdrive since it will probbably be named wrong.
i have bad internet so i figured out a way to bypass the uploading/downloading part while following your first guide, i'm on my first try right now :D, this comment is to expand my ego to new heights as i compliment my own intelligence xD.
note that if you download LORA the code in the downloader will say wrong file or link or whatever, but that's nonsense, it still downloads it to Gdrive, so just move it, rename and use it.
Hi Ad Mortem (interesting nickname by the way lol),
yeah you can, but if you want to upload them all in one go, you cannot do that, unless you run the model twice, or three times, etc.
ahah okay! lol
Quine, for real
are you an AI Laura?
LOL
A cute girl clarify my doughts thanks u buddy
Hello. I uploaded three models into the google drive folder, did the exact steps and got this specific error. What should I do in this case?
NameError Traceback (most recent call last)
in
12
13 if Redownload_the_original_model:
---> 14 with capture.capture_output() as cap:
15 rmv()
16
NameError: name 'capture' is not defined
Hi! It seems that "capture" was not imported. Are you 100% sure you run the "Install/Update AUTOMATIC1111 repo" section? That includes the below code, which is needed to run capture:
>>> from IPython.utils import capture
@@LaCarnevali I followed your advice and everything works now. Thank you very much for the support!
love how she subtedly moves past the hentai and porn categories lmfao
Hi, i've a dropbox account...is it possible to connect colab with dropbox (instead of drive)?
I think so, but you ll need to update the code