Logo to Art Tutorial. Stable Diffusion Optical Illusions.

Поделиться
HTML-код
  • Опубликовано: 10 ноя 2024

Комментарии • 146

  • @sebastiankamph
    @sebastiankamph  Год назад +4

    Get early access to videos as a Patreon supporter www.patreon.com/sebastiankamph

    • @MassimoTodaro74
      @MassimoTodaro74 Год назад

      Happy to be supporting your great work - Thank you!

    • @sebastiankamph
      @sebastiankamph  Год назад

      Thank you very much! That means a lot to me 🥰🌟

    • @Dex_1M
      @Dex_1M Год назад

      can you try the same thing for a person ?? what about 3 people ? with 3 shadows ? and what happens if you mix it with seg?

  • @lennylein
    @lennylein Год назад +25

    One hint I want to share: QR monster considers RGB #888888 as “neutral” or “transparent”. So if you’re using a black logo on a white background, it will often result in overexposed pictures that are too bright - maybe that’s what you want but that’s the effect of the white background. Try putting logos on a grey background for a more neutral process

  • @brandonMTT
    @brandonMTT Год назад

    You're a legend. I was having issues with text in my generations and the high-res fix was the EXACT thing I needed!! Also I was upscaling my images before as well and you've saved me an extra step. Thank you!!

  • @dafunkyzee
    @dafunkyzee Год назад

    Kamph, your my hero... I learn so much by replicating some experiments and exploring the ideas further.

  • @MIchaelSybi
    @MIchaelSybi 10 месяцев назад

    Amazing crerative exploration, thank you!

  • @consig1iere294
    @consig1iere294 Год назад +4

    Please do a ComfyUi tutorial on this. Thanks!

  • @jl25735
    @jl25735 Год назад +41

    My man spends 50% investigating new Stable Diffusion topics and the other 50% researching for more dad jokes 😂

    • @ElHongoVerde
      @ElHongoVerde Год назад +1

      I can assure you it's more like 25/75... 😅

    • @jonathanmogg3643
      @jonathanmogg3643 Год назад +2

      The atom dad joke is the go-to dad joke for ChatGPT.... I doubt it takes him long!

    • @sebastiankamph
      @sebastiankamph  Год назад +4

      Man, I didn't even know. I rarely use AI for the dadjokes! 😅

    • @CoconutPete
      @CoconutPete 8 месяцев назад

      there's a new sd extension out that automatically converts dad jokes into realistic images

  • @hugoruix_yt995
    @hugoruix_yt995 Год назад +1

    Just downloaded that cn model for qrs. And 10 min later you post this 😂😂 will try it out for sure!

  • @aamir3d
    @aamir3d Год назад +1

    Good one - for what you're doing, the T2I Canny XL model works well, with the Canny preprocessor enabled.

  • @jonathanmogg3643
    @jonathanmogg3643 Год назад +5

    Another good video, you provide the clearest instructions. One hour into using this model and I already got offered a job using it!

    • @sebastiankamph
      @sebastiankamph  Год назад +3

      Wow, that's fantastic! And thank you for the kind words 😊🌟

  • @grahaminatorca85
    @grahaminatorca85 Год назад

    Thank you very much for the tutorials and styles.

  • @Ayyouboss
    @Ayyouboss Год назад +1

    That dad-joke really got me. Also everyone knows Atoms got made up aswell though ;)

  • @deMuseo_cl
    @deMuseo_cl 11 месяцев назад

    This is so awesome and well explained. Thank you!

    • @sebastiankamph
      @sebastiankamph  11 месяцев назад

      Thank you very much, happy to help :)🌟😊

  • @romainalbers
    @romainalbers 5 месяцев назад

    Would be awsome if you did an updated video of this but with ComfyUI! :)

  • @shortsLLC
    @shortsLLC Год назад +2

    big thanks to you, i would really appreciate a video to how to do this with comfyui

  • @Rwae3aalameia
    @Rwae3aalameia 5 месяцев назад

    Is there a dimension choice?

  • @dirkzee1
    @dirkzee1 Год назад

    Great job Sebastian. Thank you so much

  • @techviking23
    @techviking23 Год назад

    Amazing tutorial!

  • @jaydez_
    @jaydez_ 11 месяцев назад

    amazing, thank you!

  • @Deminii
    @Deminii 9 месяцев назад

    SD have nice options to do list of prompts like u have, can we do something like that in comfy UI? ready prompts and only get freom list?

  • @Rasukix
    @Rasukix Год назад +2

    great video, quick question, why do we need the yaml files as well as the regular one?

  • @TomNorwell-f1e
    @TomNorwell-f1e Год назад +1

    @sebastiankamph hey man - how do you add styles to sd? Like your cinematic style? Thanks

  • @TheBlackBaku14
    @TheBlackBaku14 Год назад +1

    Amazing idea love it !

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      Thank you!

    • @TheBlackBaku14
      @TheBlackBaku14 5 месяцев назад

      @@sebastiankamph I would love to see a new version of this tuto but with ComfyUI ! What would you do as a workflow :O

    • @sebastiankamph
      @sebastiankamph  5 месяцев назад

      @@TheBlackBaku14 I would think it's exactly the same, just built in comfy

  • @OrangeJu1cee
    @OrangeJu1cee Год назад

    Love this, thanks for sharing

  • @dilboteabaggins
    @dilboteabaggins Год назад +1

    Have you had any luck doing this was img to img? It'd be neat to be able to take an existing photo and then drop a logo into it.

  • @TP3334
    @TP3334 Год назад

    thank you!! One thing i didnt understand, what has the qr code model has to do with any of this?

  • @HuexLabs
    @HuexLabs Месяц назад

    Hey man. I don't see the "controlnet" folder under extensions folder. I only have "put extensions here" .txt file. I was watching your video about installing it, I didn't installed any of the models from Civitai though, because I wanted to do only the logo task you show here. Is it necessary to install some models from CivitAI to go forward with it? Best!!

  • @Gr3nblatt
    @Gr3nblatt Год назад +2

    So far I did similar experiments with the lineart controlnet. In which areas do you think the qr code monster model creates better results?

  • @Drew_pew_pew_pew
    @Drew_pew_pew_pew Год назад

    Thanks man, used this tutorial to create an epic metal t-shirt. black and white skull in controlnet, prompt: lost city of atlantis and one of your styles.

  • @gabriellekamph
    @gabriellekamph Год назад

    Love this! 😍

  • @CultureCasino
    @CultureCasino Год назад

    LOVE IT

  • @coloryvr
    @coloryvr Год назад +1

    ....again: helpful and inspiring! Big FANX

  • @marcoatmac
    @marcoatmac 11 месяцев назад

    Great, thank you very much for this tutorial! But What is "cinematic" settings?

    • @sebastiankamph
      @sebastiankamph  11 месяцев назад +1

      Just some of my preset styles I use

  • @MrTheraseus
    @MrTheraseus Месяц назад

    hey there, has there been any updates to achieve this effect with comfy UI? or we still need to use stable diffusion for this controller model?

  • @khaifunk156
    @khaifunk156 11 месяцев назад

    Can we use this in ComfyUI and how to create it ? Thank you sir

  • @Mocorn
    @Mocorn Год назад

    I've done this exact thing with the older controlnet models before. What is the QR model doing differently here?

    • @aamir3d
      @aamir3d Год назад +1

      I'm trying to understand that myself. The QR model was trained specifically for QR codes, but I find using LineArt (with invert) or Canny also works.

  • @BrandosLounge
    @BrandosLounge Год назад +1

    what's the best way to make the black and white images? can we do that in sd?

    • @Fiqure242
      @Fiqure242 Год назад +1

      yes, grab a coloring book check point or lora.

    • @BrandosLounge
      @BrandosLounge Год назад

      Thanks for the tip @@Fiqure242

    • @Rasukix
      @Rasukix Год назад

      just use monochrome or greyscale in positive prompt

    • @Rasukix
      @Rasukix Год назад +1

      unless you mean for the logos, in which case just use photopea and the paint bucket tool

  • @alexling2150
    @alexling2150 Год назад +1

    Can I do this in image 2 image tab instead? For example, I found a really cool scenery wallpaper and I want to use this to get an apple logo on it. Is this possible? Thanks as always

    • @sebastiankamph
      @sebastiankamph  Год назад

      You can get a similarly themed result but not as powerful imo.

  • @ArunSingh-ld1mu
    @ArunSingh-ld1mu Год назад +1

    You said "twitter"... :-) I think you meant "X"?!?!?! LOL JK! 🙂 I think we are all conditioned to Twitter still!!! Very informative video as well!! youre a goat!

  • @LawsOfMarketing
    @LawsOfMarketing Год назад

    Doesn't seem to work, my stable diffusion checkpoint is the control_vlp qr code monster tag. Is this correct? if not, what other checkpoint should be there instead? It gives me an error message of 'runtime error 'LayerNorKernellmpl' not implemented for 'half'. What does this mean?

    • @sebastiankamph
      @sebastiankamph  Год назад

      QR Code Monster is for ControlNet. You need a regular SD model as your base model

    • @aamir3d
      @aamir3d Год назад

      Use a SD base 1.5 model, and within Controlnet, select the QR Monster.
      You'll need to use the Controlnet models within a Controlnet/models folder in the extensions, not directly as a Stable diffusion checkpoint.
      Also note You can't use the 1.5 Controlnet models with SDXL. You'll need to download the T2I or other XL Controlnet models.

  • @Cavemanmediahousebiz2
    @Cavemanmediahousebiz2 Год назад

    In my extensions folder I don't have the folders ' sd-webui-controlnet ' like you do. I have downloaded the two files on hugging face. Am I missing something? Do I just create sd-webui-controlnet and models and drag the files into there?

    • @sebastiankamph
      @sebastiankamph  Год назад

      You need to install the extension ControlNet. This guide doesn't show that, check my ControlNet install guides.

  • @quizkraftstudio
    @quizkraftstudio 6 месяцев назад

    what's the best way to get a 16:9 aspect ratio? It does not work with this setup. Thanks.

  • @shinocg6159
    @shinocg6159 Год назад

    hey man, i have seen multiple SD webui users who used some kind of text-correction/text-suggestion when writing the prompt text. i can find how to enable that feature anywhere. can you help

  • @siddhid4746
    @siddhid4746 Год назад

    I’m not getting any image on the right side. Just grey patch.. any idea how to solve that?

  • @7210LN
    @7210LN Год назад

    good👍

  • @CoconutPete
    @CoconutPete 8 месяцев назад

    does anyone know what the purpose of the yaml files are?

  • @awesomesoftwares1975
    @awesomesoftwares1975 Год назад

    it says ModuleNotFoundError: No module named 'cldm' , what should I do?

  • @dthSinthoras
    @dthSinthoras Год назад

    Love this! But for my tries the Logos are very often very faint, almost blurry. Any idea how to fix this? Is this more likely about the model, the vae, cfg, steps, control weight,...?

    • @sebastiankamph
      @sebastiankamph  Год назад

      Raise the weight and use hires fix. I talk about this in the video.

  • @Braveheart201
    @Braveheart201 Год назад

    Any reason why you downloaded the V1 and not the newer V2 files? The V2 folder is kinda hidden, its a subfolder on huggingface

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      Oh, that was just a mistake when recording. Feel free to get the v2.

    • @Braveheart201
      @Braveheart201 Год назад

      awesome, thanks!@@sebastiankamph

  • @riiveyt
    @riiveyt Год назад

    The prompt styles are not free in your patreon account. How do we get them free?

    • @sebastiankamph
      @sebastiankamph  11 месяцев назад

      These are now changed to Patreon subscriber only

  • @mightypensword
    @mightypensword 10 месяцев назад

    I've followed your steps, but there is not a hint of my logo in any of my attempts. Even when the blurry image first appears, my logo is not there at all. Ideas? The only thing I'm missing from what you did is the "cinematic" option. I have no options when I click that box

    • @mightypensword
      @mightypensword 10 месяцев назад

      When I make the control weight 2, and select "ControlNet is more important", I get just a hint of my logo shape coming through. Do I need to weight the logo more, somehow? Is the fact that my graphics card is only 6gb a factor?

    • @sebastiankamph
      @sebastiankamph  10 месяцев назад

      Did you go through hires fix as well?

  • @mcfcmandem
    @mcfcmandem 10 месяцев назад

    Would this be possible on a video?

  • @chouaibphenix1082
    @chouaibphenix1082 Год назад

    How do I open the program? I have followed all the steps But I couldn't find anything in the file extension 😢 I did everything But I can't find the same files as yours I ask for help. It is two o'clock in the morning, but to no avail

    • @sebastiankamph
      @sebastiankamph  Год назад

      Check out my Stable diffusion for beginner's full guide 2023 to get you running

  • @not_riley
    @not_riley 10 месяцев назад

    hi, you said in the video the prompt styles are available for free but when i followed the link it requires you to be a paid subscriber.

  • @andy111007
    @andy111007 Год назад

    Greetings Sebastian, Thanks for the amazing video/Where can i get a list of styles?

  • @shaunbass444
    @shaunbass444 Год назад

    Hey this is great, im just wondering can you link me to the Epic Realism model download. I cant seem to find a link - or maybe im just not seeing the right files i need.
    Would be a massive help thank you.

  • @lyonstyle
    @lyonstyle Год назад

    man oh man, its been a while since i touched Stable Diffusion. i think its time to get back at it 😂😂

  • @lortzo573
    @lortzo573 Год назад

    i dont have any models in the controlnet tab

  • @muharremacar5255
    @muharremacar5255 Год назад

    how can I turn an image to black and white? a face, for example...

  • @sourabh662
    @sourabh662 Год назад

    Hey brother, i followed your video about how to install this and now came to this video. You are putting files in extension folder but in my case my extension folder is empty so how can i put this file into models??? please help!!!!

    • @sebastiankamph
      @sebastiankamph  Год назад

      You can create the folders. But if you have no extensions folder, you probably have no extensions. Go into extensions tab and install ControlNet first.

  • @lortzo573
    @lortzo573 Год назад

    i've been trying to make qr codes with this but i didnt manage to make even one scannable code, do u have tutorial on that?

    • @sebastiankamph
      @sebastiankamph  Год назад

      Of course ruclips.net/video/HOY5J9UT_lY/видео.html

  • @Effetmr
    @Effetmr Год назад

    hello Kamph, thank you for all these useful videos.
    I tried your method and did everything like you but i am getting this error : RuntimeError: mat1 and mat2 shapes cannot be multiplied (154x2048 and 768x320)
    how can i fix this please ?

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      Can't mix sdxl and 1.5

    • @Effetmr
      @Effetmr Год назад

      @@sebastiankamph thank you it worked perfectly ! keep the good videos flowing we getting addict hh
      i suugest a topic : how to use deforum on InvokeAI ?

  • @Rasukix
    @Rasukix Год назад

    quick question, what does highres steps do?

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      The number of steps for your highres pass, ie the 2nd/final pass of the generation.

    • @Rasukix
      @Rasukix Год назад

      if we leave it at 0, it matches the normal steps tho@@sebastiankamph

  • @andrekerygma
    @andrekerygma Год назад

    I think the discord link is not working... =/

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      Yeah, lost my boosts. Check in channel profile

  • @20xd6
    @20xd6 Год назад

    How would you go about making it 4k?
    ultimate sd upscale with the cnet model?

    • @sebastiankamph
      @sebastiankamph  Год назад

      Probably just a regular upscale. I don't the cnet tile upscale is warranted, but can be used for better results.

  • @tiffanyw3794
    @tiffanyw3794 Год назад

    Can this be done in sdxl

    • @sebastiankamph
      @sebastiankamph  Год назад

      Yes, but not with that model. Try the Canny sdxl model. It won't be exactly the same, but similar.

  • @Rasukix
    @Rasukix Год назад

    does the input photo require a background/have to be the same dimensions as the output generation also? (e.g. square input image if outputting 512x512)

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      No, can use pixel perfect, but that will change where it ends up on your final image.

  • @arskiz
    @arskiz Год назад

    why it took so long for you to make those images? for me it takes like 2seconds to create one lol

  • @arunpriyan524
    @arunpriyan524 Год назад

    OutOfMemoryError: CUDA out of memory. Tried to allocate 512.00 MiB (GPU 0; 4.00 GiB total capacity; 3.02 GiB already allocated; 0 bytes free; 3.09 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
    I got this error. What should I do next?

    • @sebastiankamph
      @sebastiankamph  Год назад

      Buy a new GPU. Or load (in your webui-user.bat file) with arguments in --lowvram

  • @hanshans4814
    @hanshans4814 6 месяцев назад

    I don't know if you'll ever read this but i'm from germany and have a Samsung Galaxy S10+(released in 2019) and i tried downloading some pictures and messing around with it. So i downloaded three illusion pictures and wanted to make a slide with it in tik tok. But while i was creating the photoslide my S10+ went nuts, started flickering on and off, not responding, green flickering lines started to appear but luckily it stopped when i exited tik tok in that one moment where the screen was on. My heart almost stopped, i thought i just lost my phone in that moment i wanted to cry. Please could you tell me why this happened to me?

    • @sebastiankamph
      @sebastiankamph  6 месяцев назад

      Sorry to hear that. I have no idea. Most likely one of 2 things. You had a software error, ie an app that was broken, or your phone is starting to give up.

  • @hdslave
    @hdslave Год назад

    can we have that cinematic style preset?

  • @lithium534
    @lithium534 Год назад

    LOL! Did you get a letter from apple for your other thumbnail? Or is Mr Beast just everyone's milk cow.

    • @sebastiankamph
      @sebastiankamph  Год назад

      Just trying to find what people will click on 😁

  • @nousername7817
    @nousername7817 Год назад

    1a wie immer. kaum war das Video draußen wurde es auch schon von anderen kopiert. Kann man ja auch als Kompliment nehmen. danke.

    • @sebastiankamph
      @sebastiankamph  Год назад

      Thank you! It's hard work as a smaller creator 😊🌟

  • @tedabghari
    @tedabghari Год назад +1

    Sebastien, I would love to see you shifting gears a little bit and starting to make videos of WarpFusion.

    • @sebastiankamph
      @sebastiankamph  Год назад

      Not a bad idea

    • @tedabghari
      @tedabghari Год назад +1

      @@sebastiankamph there are so many Contant creator’s making videos on stable diffusion and technology is what it is. There is not much more you can add. But Warp has so many different possibility’s of settings, and there is no Contant creator that is making advance tutorials on how to warp only the basics.

  • @erikjohnson9112
    @erikjohnson9112 Год назад

    Atoms make up everything has become an AI joke. Artificial Dad Intelligence?

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      I've learned that now, but I didn't know that before 😅. ADI surely must be a thing, no? 😁

    • @erikjohnson9112
      @erikjohnson9112 Год назад

      @@sebastiankamph I noticed this while watch way too many videos about running LLMs locally. A common task was to write a simple web app that tells jokes when you press a button. It seems like half the time the joke would be this one ("atoms make up everything"). Btw, running local LLMs is getting a lot better.

  • @trinibaduk9012
    @trinibaduk9012 Год назад

    I don’t think u know how illusions work, your suppose to lose the image and squint your eyes to see it, that first twitter was good

    • @sebastiankamph
      @sebastiankamph  Год назад

      Feel free to lower the controlnet weight and you will see less of the image :)

  • @jonmichaelgalindo
    @jonmichaelgalindo Год назад

    So, is the old Twitter logo homeless now? Is it just abandoned? (I'm guessing probably not, because lawyers.)

    • @sebastiankamph
      @sebastiankamph  Год назад +1

      In the philosophical sense, I guess. From a legal standpoint, they very much still own it 😅

  • @IlRincreTeam
    @IlRincreTeam Год назад

    Now this dadjoke was fire (it caused mayhem in Hiroshima)

  • @MrSongib
    @MrSongib Год назад

    Dead birb. 😥

  • @AaliDGr8
    @AaliDGr8 Год назад

    huhh what a shit pitty maaaan why my side qr monster sucks i just type 13 words long text then make it png with white background and black text then after that i try none then linerart 2nd one u show in control net then i pick always qr monster in CN model tried dreamshaper model but my results always super crap not even illusion hpnd why dont u do it with text long text plzzzz

    • @sebastiankamph
      @sebastiankamph  Год назад

      The words need to be big enough. If they're small you can try to increase the weight to the max.

    • @AaliDGr8
      @AaliDGr8 5 месяцев назад

      @@sebastiankamph plz plz plz make this same tutorial more easy in/for Defooocus plzzz ??? hmm