Stable Diffusion Realistic AI Consistent Character (Instant Method Without Training)
HTML-код
- Опубликовано: 14 май 2024
- 🔔 Subscribe for AIconomist 🔔
👑 Get Canva Pro Lifetime Access for $9.99!! ➡️ canvadeals.com
1️⃣ Stable Diffusion Checkpoint Model:
epiCRealism ➤ civitai.com/models/25694?mode...
2️⃣ LORA:
epiCRealismHelper ➤ civitai.com/models/110334?mod...
3️⃣ Extensions:
Roop ➤ github.com/s0md3v/sd-webui-roop
Ultimate SD upscale ➤ github.com/Coyote-A/ultimate-...
4️⃣Upscaler:
4x-NMKD Superscale➤ huggingface.co/gemasai/4x_NMK...
5️⃣ epiCRealism Video ➤ • Stable Diffusion Best ...
6️⃣ Controlnet Tutorial Video ➤ • How to install and use...
🖥️ my PC setup:
GPU - amzn.to/3PAbjP6
CPU - amzn.to/4cngN9A
RAM - amzn.to/494I5ig
SSD Storage - amzn.to/3x739r4
Prebuilds PC for Generative AI - amzn.to/3TQDTOQ
For Business: info@aiconomist.cc
Music:
Opus One by Audionautix is licensed under a Creative Commons Attribution 4.0 license. creativecommons.org/licenses/...
Artist: audionautix.com/
Thank you, bro! Keep going) The content is simple and short. And exactly what I expected))
I like how the tutorials are short and to the point and not making it longer solely for the ad revenue
Wowwww!!!! Wowwww!!! Thank you. BEST tutorial ever. You are the BEST!! ❤
You can simply use the "resize by" option instead of the "aspect ratio calculator", it works faster :)
Thank you really much, this is the best tutorial I've seen, really helpful bro ! I just followed you
Thanks for the support!
thanks man you deserver more subs
Thank you for this incredible tutorial, my like and whattime! 100%
Thank you so much for appreciating the tutorial! Your support truly means a lot. Stay tuned for more content like this. there's plenty more to come! 💡🙌
@@Aiconomist 💜💜💜
Great work, to the point :)
Excellent because the lore I don't do good, I will test soon
Thank you for the Video. is there a Checkpoint Model for SDXL 1.0 too? thx
Thank you always. I succeeded in changing my face through reactor, is there a way to change my outfit and hairstyle naturally?
Thanks for sharing this, I have subscribed to your channel. I was wondering if this technique is possible to do using Comfy UI?
Très bien expliqué, abonné !
Merci!
Do you know how I can upload multiple prompts (for instance for difference scenes at once and also select a certain amount of cues)?
Would be interested to know on how to have multiple photos of the same face so it covers all the angles and then use that ''trained'' version for this.
Nice add-on!
Nice! Does this work on Radeon Amd Graphics and Ryzen 5?
Mine is showing TypeError: Script.run() missing 18 required positional arguments: '_', 'tile_width', 'tile_height', 'mask_blur', 'padding', 'seams_fix_width', 'seams_fix_denoise', 'seams_fix_padding', 'upscaler_index', 'save_upscaled_image', 'redraw_mode', 'save_seams_fix_image', 'seams_fix_mask_blur', 'seams_fix_type', 'target_size_type', 'custom_width', 'custom_height', and 'custom_scale'
this error. What can I do now?
Where did you gain so much knowledge about such in depth AI techniques? You're level of experience is insane.
Thank you for your kind words! i gained much of my knowledge from amazing individuals on RUclips and various online resources. Their expertise and willingness to share inspired me to start this channel and share what i've learned with you 😊
@vo6pk he's an AI, GAI 😂😅
Thank you, but I got this error:
RuntimeError: Input type (c10::Half) and bias type (float) should be the same
Do you have any idea how to fix it?
Loved your videos! Can you also make a video about face replacement for videos ? Thanks.
Yes, soon
@@Aiconomist still waiting :) Also do you know the solution for this video I get as an output only black paint where I do inpaint ?
How about the image like line art or something like that, can i generate/swap my photo on it?
Hi! can you do tutorial about all sampling methods in Stable Diffusion?
Absolutely i think it's a fantastic topic
thanks
You're welcome!
Can you do this with Fooocus or only in SD?
Thank you~
hello, I watched your video that talking about creating paint book on A1111. I installed every thing that you said. But when I click on generate button its says:
''OutOfMemoryError: CUDA out of memory. Tried to allocate 114.00 MiB (GPU 0; 1.95 GiB total capacity; 1.68 GiB already allocated; 58.25 MiB free; 1.73 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF ''
so what shoud ı do ?
my computer operating system is Linux.
Que grafica tienes? por que si tienes menos de 6Gb de vram tendrás muchos problemas.
*What's your device configuration?*
"The face has been seamlessly replaced with no issues at the edges."
And this is why os many people say AI images suck, you don't see what's wrong with it.
The color shading and grain is different from the rest of the image, and the face doesn't match the head. The new face is narrower making it look like she has a sack below her cheek.
Where is the SA folder? As in "Put into Models folder in the SA folder" after downloading the 4x NMKD Superscale Upscaler
Hi man You ate Amazing, great tutorial. Work it in forge?
Sorry for been late, use a new extension called "Reactor"
I have 4GB nvidia graphic card and 8GB of RAM in my computer. will it work for me as i previously tried fooocus and comfy and both shows out memory error.
Your sound is like mkbhd :)
can this apply to 3d animation style character for children book?
How do you change hair tho?
can you run this program on a MacBook pro M2 2023 ?
hey, great tutorial... but i ran into some issues with the roop, i have installed it but it doesnt show up
Hi Glad you like it, However you can use a Roop alternative Called Reactor
Do you need a powerful computer to do this locally, so a powerful enough video card? Because I have a laptop and I don't know if it can generate such good quality.
you do, you can use it on a lower end pc but it's going to take a loooot of time generating pictures, especially if you want to do something like this
thank you so much for the video. do you know why roop doesn't show up under control net? i have it downloaded and did it the exact same way as you.
Find where your python 3.10 install is located (probably C:\Users\*****YOUR_USERNAME_HERE****\AppData\Local\Programs\Python\Python310\), go to lib > site-packages. Copy everything from inside that folder to your SD > venv > lib > site-packages folder and overwrite everything and restart SD. you should now have a working roop.
it was because roop was discontinued you have to use an alternative
same for me
Best
We need this for Fooocus
Can you guys tell me how to install all these on Runpod?
I like the video. Keep them coming. But... Roop is abandonware now. There are better options...
Absolutely right! there's Reactor, and I've also recently created a video about IPAdapter.
im having "AttributeError: 'NoneType' object has no attribute 'model'" error when i press preview
i am using stable diffusion in google colab but it keeps on disconnecting any solution to this problem
Hi, Im kind of new in this Im learning how this work, Can I use it if I have a Mac computer?
'm sorry, but I'm not a Mac user. However, I've heard that the M1 chipset works really well with A1111.
roop is discontinued, any other suggestions
This is for Mac or Windows ?
Do you have to pay for any of this?
Is it possible to apply clothes and have them look exactly the same when they are slightly different?
Yes, you can check my last video.
i will stick with my lora tnx as i replace the body shape too not just face
Hello, where is the link to the extension?
Sorry for that i just added the links in the description
@Aiconomist hi can you help me please? the Ultimate SD Upscaler doesn't show up I have followed all steps and even reinstalled a few times wont seem to show up for me
It works but you can still see a difference in the face and the rest of the body, it looks somewhat plastic.
People will just think it's filter or smth, from my experience Instagram is mostly filled with brain-dead viewers
50 IQ hungry simps won’t notice
is it possible for mac?
I don’t understand the purpose of inpainting and openpose in the workflow. I use roop all the time without those extra steps and haven’t found an issue with it.
If you continue to use roop. Soon you will be realized the reason of using inpaint openpose and lora
+ inpaint part that showed in this video can be done without any word in a prompt, without openpose and with zero denoise strenght, and it’s not clear why he set 50 only mask padding pixel and didn't choose Only masked. lol
@@abzaccc I agree, there are many steps he does that aren't needed. All insightface swap extensions do not need manual inpainting, it already does this better than a human can and in miliseconds
I can't get the roop extension to display in the UI, please could you help? there are many people on the net who have this problem
Same problem
Same problem
Same
same
same
Hi, nice video. I did everything and everything runs fine, but it doens't appear the tab for the roop extension. Any idea why?
Already solved it. If any of you have the same issue make sure to install every requirement that is in the \stable-diffusion-webui\extensions\sd-webui-roop
equirements.txt file. It is not mentioned in the video so if it's not working for you, you should install them. Good luck guys
the face swap is not the problem its a problem for me to find a person with similar hairs to my ai
is there a way to train midjourny so it will create always similar looking persons?
Suggest me any site😢
Is this narrated by an AI Marques Brownlee?
Roop is installed but refuses to populate in the WebUI. Starting to think I should just give up on AI
Roop is discontinued search and Install another Faceswap extension called (Reactor)
I have roop problem on my computer it didn't show up. I deleted my everything and started over, controlnet, roop, adetailer... everything worked, but roop
I'm sorry to hear you're having trouble with Roop. It's worth noting that Roop is outdated, and the project isn't being actively updated. however, there's a new and promising solution for fast face swapping called "Reactor." i'll be creating a video about it soon 😊
@@Aiconomist I've following you then. Thnx
2:08 When you want to change the value of "Only Masked Padding, Pixels" you want to select "ONLY MASKED" which to the left of that slider, otherwise the value you set with the slider has no effect.
What "ONLY MASKED" does, is re-drawing just the masked pixed, and the padding pixels you've selected with the slider which extends out around what you've masked.
What "WHOLE PICTURE" does, is re-drawing the whole picture but only edits what has been masked.
ONLY MASKED is good when making small, fine and detailed changes to eyes, buttons, jewels, etc where you want the most focus. When done correctly, you will see how the preview image zooms in to focus on the masked area.
WHOLE PICTURE is good when you want to make larger changes, like the color of a shirt, hair, the shape of a hat.
Both of these can also be combined with "LATENT NOISE", if you want to add something new to the image. If you keep "ORIGINAL" selected it will just use what noise is already in the image.
2:26 No, click the "RESIZE BY" button to skip the calculation.
4:55 No. Ultimate SD Upscale produces a grid which is highly notable when changing the gamma and saturation, do not use Ultimate SD Upscale unless you know for a fact you're too short on VRAM to upscale to a somewhat standard 1440p image. If you have 8GB or more, don't use this upscaler. Either increase steps and resolution while in your standard IMG2IMG tab, or throw the image to Extras and pick a non-grid upscaler like "Nearest" or "ESRGAN_4X."
but how do you make consistent characters with the same clothes and in the same location?
creating consistent characters with the same clothes and in the same location often involves techniques like inpainting and other photo editing methods. i'll work on simplifying the process and cover it in a future video
@@Aiconomist thats cool, thanks bro
Have installed the roop extension but there is no option to use it, there is nothing under controlnet, unable to install insightface dependency
same
same
ROOP is dead, use REACTOR instead
Or better, Faceswaplab.
@@duskairable haven’t tried it yet, can’t get my hands on it. Is it fast enough for RTX3060?
@@duskairablewhat is this
@@Max3DDesignuse cloud lab instead
@@tacorevenge87 will try, thanks
new era catfishing is gonna be WILD!
any guy who sends money to someone they never met irl is down bad ol
I follow your steps but roop does not show up for me. HELP!
same problem with me
Not the AI MKBHD voice 😂
Why are there like 20 applications that I need to download
where is that calculator?
my laptop doesn not have a graphics card, can i run this program ?
Yes it will but you will have to edit the webui-user batch file and and some lines . Will definitely take longer to generate images with no gpu
Why your voice sounds like MKBHD????
this sounds like mkbhd
is it MKBHD's voice?
Hahaha a MarkASS Brownlee cloned voice. Good call. Better than “Anthony” everyone is using.
Please "NameError: name 'xformers' is not defined"
So i can put my face on a godfather escene
roop is no longer working and has bugs in its installation as of JANUARY 02, 2024.
i got it working, needs a bit of tech understanding
A hour later I finely got Roop to work.
didnt work for me. Any idea on how to get it working?
I read the txt file and make sure you have all those files installed. @@mrprimogems
roop doesnt show up
If the Roop extension isn't showing up, please ensure that the "sd-webui-roop" folder is located inside the "extensions" directory. Additionally, make sure to close and restart Automatic 1111 each time you add a new extension.
still not showing up@@Aiconomist
You can use Reactor instead, however i made a new video on how to make a consistent instagram model. check it out!@@ethicalchatbot7229
I swear the person speaking is Marcus Brownlee. If you read the comment, good job but mask off ahaha
Err what just happened? This is too much nerding for one day
Why do you sound like MKBHD
Using mkhd’s voice is crazy lmao
Wow you voice cloned MKBHD
Is it paid or free ?
I find it scary
roop is dead now :(
Not very good result at 4:03. Look at the left cheek and neck, a lot of fat there considering the replaced face is narrower
i understand nothing
Noooooooo. Enough.
How can I prove your voice is AI generated? 😂
The problem with this is that only white men and white women can be depicted as high res realistic human like renders. Any other ethnic group and the AI has trouble. That's because the data it was training on did not adequately include other groups and therefore was not trained on.
AI (programmed to do) not create those pics they allready input. we allready had reslut if do photoshop. and i want see peoples who see imahe. thats me BUT angelina jolie face. start sue chatgpt and all stop. bcoz they dont have right for image
ai is 💩💩
imagine caring about ai in 2024
This voice sounds familiar, @marquesbrownlee
Hi, can i use this with my intel iris xe ?
definetly no