Picture Yourself with Stable Diffusion
Vložit
- čas přidán 28. 06. 2024
- A quick tutorial about creating your own Stable Diffusion model in four easy steps, using just a few photos of yourself and training them with Dreambooth.
All you need to do is running the free fast-Dreambooth notebook by theLastBen, using Google Colab, to create your Custom model and then import it into a local Stable Diffusion app, like Automatic1111.
To watch some more great music videos created with StableDiffusion, Unreal Engine and Blender, visit our CZcams music channel:
/ @-vero-
----------------------------
Chapters:
00:00 Intro
00:22 Four easy steps to create your Stable Diffusion model
01:59 Preparing & uploading your photos
03:27 Training your model with Dreambooth
06:57 Downloading and importing your model into Automatic1111
07:51 How to use your model
08:25 Sample Prompts
Workflow:
----------------
Take 15 portrait photos (more or less),
good lighting and contrast, no hard shadows, face fully visible
no other people, different backgrounds, different angles, different expressions
Resize the photos to 512 x 512 pixel with Birme.net
www.birme.net
Rename the 512 x 512 px photos, using a unique expression, which will be the trigger word for using it in prompts. All the photos should have the same name plus a sequence number (veroLab (1).png, veroLab (2).png ... veroLab (15).png). Leave a space between the keyword and the sequence number.
If you want to train several people into one model, rename the photos of each person with a different keyword (sophieLab (1).png, sophieLab (2).png, ... geraldLab (1).png, geraldLab (2).png)
Upload the renamed images into a new folder on your Google Drive. (If you haven't got a Google Drive, create a free Gmail address on Google and log into your browser first)
drive.google.com/drive
Open the lastBen fast-Dreambooth notebook on Google Colab to train your model (detailed explanation in the video)
colab.research.google.com/git...
Github Page: github.com/TheLastBen/fast-st...
Download the trained model from your Google Drive and import it into your StableDiffusion App. I would recommend Automatic1111 but there are also alternatives available
Local installation guide for Automatic1111 on a Windows-PC:
stable-diffusion-art.com/inst...
and on a Mac with Apple Silicon:
stable-diffusion-art.com/inst...
other SD Apps:
DiffusionBee (only for Mac) diffusionbee.com/
InvokeAI: github.com/invoke-ai/InvokeAI
EasyDiffusion (Windows and Linux): stable-diffusion-ui.github.io
Downloading custom models for Stable Diffusion:
civitai.com/
-----------------------------------------
Some great CZcams channels, covering Stable Diffusion:
/ @oliviosarikas
/ @nerdyrodent
/ @sebastiankamph
/ @promptmuse
#stablediffusion #automatic1111 #dreambooth - Krátké a kreslené filmy
It really works. It will be very helpful for those who do not have a powerful graphics card but want to create a ckpt model. I've been looking for something like this for a few days. Thanks.👌
Works like a charm. Thanks!
This is an excellent guide!. I followed this step by step and it worked! can't thank you enough for the clarity of this tutorial. honeslty I think your channel is underrated. If I had one request it would be to ask for that prompt in the final section in your description, or perhaps prompt styles/guidelines that would work best with this trained model. Thanks again!
Absolutely brilliant!! I can't thank you enough. I am now addicted to generating images in Stable Diffusion from self-models following this tutorial. So fantastic I could hug you. 😃
Whose pictures do you usually create? Your crushes or gf? Or actresses? :p
Thank you so much fow this easy and detailed tutorial! im a total noob with computers and very new to stable diffusion and until this video i was just confused and intimidated because i dot know any of these words and couldt follow. but this time i made it and it worked and tbh it was so much easier than expected... thx a lot!!!
It really works. Thanks
Thanks!
REALLY NICE! tHANKS A LOt
I am subscribed. Can you make a tutorial on how to train a model locally using your own computer ?
Thank you for a very easy to follow guide! Do you know a good guide that shows how to do the same with google collab but creating a Lora instead?
Terimakasih orang baik ! ❤
Good results. So don't you need text files for the images ?
Any solution for
AttributeError: 'NoneType' object has no attribute 'pop'
after try load model?
Good
thank you, can traine kohya-LoRA-dreambooth?
Tried running this but errored out... in your video you have the safetensors check box, but the link in your comment section does not show it.
changing setting sd_model_checkpoint to Jen.ckpt [afdbcad2ef]: AttributeError
Traceback (most recent call last):
File "C:\SD\stable-diffusion-webui\modules\options.py", line 146, in set
option.onchange()
File "C:\SD\stable-diffusion-webui\modules\call_queue.py", line 13, in f
res = func(*args, **kwargs)
File "C:\SD\stable-diffusion-webui\modules\initialize_util.py", line 174, in
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: sd_models.reload_model_weights()), call=False)
File "C:\SD\stable-diffusion-webui\modules\sd_models.py", line 773, in reload_model_weights
state_dict = get_checkpoint_state_dict(checkpoint_info, timer)
File "C:\SD\stable-diffusion-webui\modules\sd_models.py", line 330, in get_checkpoint_state_dict
res = read_state_dict(checkpoint_info.filename)
File "C:\SD\stable-diffusion-webui\modules\sd_models.py", line 314, in read_state_dict
sd = get_state_dict_from_checkpoint(pl_sd)
File "C:\SD\stable-diffusion-webui\modules\sd_models.py", line 253, in get_state_dict_from_checkpoint
pl_sd = pl_sd.pop("state_dict", pl_sd)
AttributeError: 'NoneType' object has no attribute 'pop'
I have the same problem. Have you find out how to fix it ?
the safetensors check mark doesn't show up and i am unable to connect to a model. I am stuck at this point. thanks
hi can you do lora version?
Has something changed with this process? It seems like the "Safetensors" checkbox in the Model Download section is gone.
bro you go step by step and then at 8:13 you just quickly selected multiple specific options and styles that you did not describe in the video, and did not describe how to get those styles for 1111 for first time users like me. Gave me surprisingly different results when I first tried it myself - I appreciate the rest of the video though and will continue researching how to make better portrait photos
Only faces? Or can we zoom out?
does anyone know how to fix "Expected all tensors to be on the same device, but found at least two devices"? I followed the tutorial but alway got this error.
How can i download stable diffusion?
untimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument index in method wrapper_CUDA__index_select) Got this error
Thanks, but when i try to load the model (.ckpt) it errors out... tried twice because i loaded the images with .jpg first and it error out. I changed to .png but still error...
What I notice is that "SafeTensor" option is missing in the google colab.
EDIT: NVM. I updated SD to latest version and it all worked. Thanks for the great tutorial!
How'd u update your SD?
How'd u update your SD?
5:03 need it for??
Why does it not work anymore? with xformers
pay me, I would teach you how to make it work successfully
@@HolyDevil2402 🤣
5:01 in civitai you said that need that information later.......I don t see you saing about it. Ah one more thing my colab don t have safetensors check>>?? give me error
yeah im stuck there too
As I understand it, this method doesn't work anymore?
think so. wasted an evening
I couldn't do it. Yes, seems not working
Works fine!
Hi
error
It's a great tutorial but outdated already. The safetensors checkbox is gone and without it SD thinks it's unsafe and errors. It say you can code the change to run it but for unsafe checkpoints but that would be really stupid. Google designers must be the left over mental midgets from Microsoft.
but you created a model, i need a lora