Quick Tutorial for DISCO DIFFUSION "init_image" (Steampunk the Mandalorian helmet!)
Вставка
- Опубліковано 12 вер 2024
- Well, I had to do it!
So here's a simple way to use your own image and Disco Diffusion them.
It's a superb tool for digital concept artists...
Have fun!
/ quick_eyed_sky if you want to support my work - and get tips (and prompts).
you're doing mankind a service. thanks for making these inspiring and easy to follow videos!
I really love how you not just show the how but also tell the why, keep making banger tutorials
yeah the lack of fluff on this was amazing. answered every question i had, thank you!
I'm diving deep into this and have some fairly decent results from your first tutorial! Please keep 'em coming! You have opened a Pandora's box for me lol
Great tutorial! 👌 Never stop doing these
Omg this video is game changing for me. I am a traditional artist and unfortunately a perfectionist. So the fact that I haven't been able to get disco diffusion to match my vision has been driving me insane to the point where I'm like fine I'll just paint it myself. I love the images it produces, but they have never been exactly what I was looking for. Did not realize I could give it a reference to work from.
That's the point. DD is a wild animal, you can't tame it, it's daring and silly. You just have to tango. Using the partials is a way. Mastering init_image is another.
Glad you put all that effort into this. I'm sure most people (including me) wouldn't bother LOL.
This was the missing tutorial I needed. I couldn't figure out how to add my own images, and what was going on with changing ratios. Thanks so much! I also agree with another commenter, how to add multiple reference images for an animation would be great. Thanks!
Loving your vids on DD. It's a beast, can't wait for dall-e 2 to let me in.
Whoa! That’s so cool! Thank you so much for this information and inspiration. I have a bunch of ideas now. I’m excited to try this out.
This is unbelievably amazing !
well done! the aristocratic cat is so good
The best thing on the internet right now!!!
So amazing !! Thank for your videos
Great tutorial man ! definitly helped me alot
Thanks so much! I was stuck, because I pasted the path into image-prompt instead of init-image.
Very educational and easy to understand videos. Thank you 👌🏼
Thank you so much :D this stuff is so inspiring....any other musicians out here overrun by inspiration?
it’s really helpful! Thanks a lot!!!
You're videos are amazing!! You have helped give so much insight in this I love it :)
Your.
This is AmAzInG...(O,o), thanx alot!!
Thank you so much!
Hey JP, I appreciate the tutorials but I'm loving your music more... Where can I find your music?
Not so easy for those used to easy sliders and conventional UI. Thank you for taking the to explain this. Initial image is critical for those who want to control the composition.
Yes I understand. I hope you got good results.
Wow!! Thank you 👍👍
So I made a music video a while back, where I jokingly put in the end credits that every frame was generated by AI. I had no idea the technology now has come so far that this is actually possible.
I have not "content" folder...
cool , thanks man ! :)
Thank you for this amazing tut.
is it possible to add more than one image? or a video so it makes images on each frame???
If I delete the path of the imported picture, I can render normally, but as long as I add the picture path, I can't continue rendering. Do you know what's going on
Please tell me prompts you write for building. thanks
Please make more tutorial 🙂
Wdym the model will get freedom at 3:33. As in like if you lower it, it will ignore the model/int-image more?
Hi, why at 3:03 u changed width_height to 1152, 768?
is 1,152 just an arbritrary number for stretching the image or is there some other calculations behind it?
With init_image, your pixels must be proportional to the original image, AND multiples of 64.
you are the best .
evar.
can someone explain to me why i set 10 for init scale and the image output still the same as original? Please help me.
If you will add also a voice to you tutorials, even generated, they will receive much more views. Great channel anyway
Yeah, but no. I don't want to be like all these artificially happy youtubers. I want to make short videos you can watch in silence, straigth to the point, without voice, without endless idiotic credits and intros, without the terrible "13 minutes for success", and without any lamps in the background.
Respect 🤘
Thank you 🙏❤
I didn't understand how did you do it can you show me on discord or something ??
Good stuff! Is there a link to the other tutorial?
Hi, thank you! The tutorials are here: ua-cam.com/play/PLc9_vneTcTGXdjoCD0b_-2x3-HqsvxCZH.html if you want to make your own images.
so do I have to have an 'artstation'?
Hi, it's just a way to show your work.
what prompt at 5:08
Legal demais, como sempre
can't find content and init images, even when i've connected my drive
You mean the folders? Add them manually if you want. But they should be there.
What prompt and settings did you use to make the ship 'painted'?
Where is it in the clip?
@@QuickEyedSky 5:05-ish, ship, you made it look painted
@@MaximusEnergy "great painting of a royal sailing boat in spring by greg rutkowski, artstation" - nothing so special :-)
@@QuickEyedSky How many steps and skiup_steps? init_guidance? Thanks!!!
@@MaximusEnergy yeah, i'm trying to make a photo in a painted style, not getting good results
Thank you! Do you always use artstation in promt and why?
Almost always. It gives a "digital painting" quality. Try without and with, and you'll know.
You can totally not use artstation or any of the commonly used prompt words and get great results. They can help though.
Artstation just happens to have a ton of reference and high quality source art from very talented artists. If you put in that prompt it will most likely trend towards the most prolific artists on there, as they just have more material available to the AI to train on. Eventually everyone will be outputting that same style so it will become a bit homogeneous. (not a good thing or bad thing) If feels a bit off just being able to tap into other peoples' styles like this but I guess there's not much you can do. This AI is game changing for the art world. The more reference art or images available online of that artist, the more the AI has been able to train on.
I would experiment with artists styles that you like if you want to get specific or play around with specific artistic terminology and techniques. I really hope there is a way to input a trained set of specific "style guide" images for artists that manage to actually have a style but just want an easier way to train the model in their own style...or a specific style.
Amazing!! how do you make with the signature?
What do you mean?
@@QuickEyedSky is that sometimes it looks like a signature is floating in the image. I thought it was on purpose. But I realized not. Sometimes it generates these squiggles that actually look like a signature. :)
@@fredwolf8617 I adore every mistake it makes. Today I got 2 or 3 signatures in a batch! Very cool, right? The AI says: " I made it!".
@@QuickEyedSky yes this is it :D. Thanks for your work.
how can i unmount my drive?
Why would you do that?
@@QuickEyedSky do people have access to it? Also I did this and am not seeing an AI folder.
Any idea on a workaround for this "FileNotFoundError: [Errno 2] No such file or directory: '/content/drive/MyDrive/AI/Disco_Diffusion/models/secondary_model_imagenet_2.pth'" ??
Having this same problem. Keeps hanging up on the Diffusion and CLIP Model Settings part. Have you solved this?
@@MarisBlack nah I gave up. I tried linking to to another gdrive and nothing changed.
I have the same problem...
/content/drive/MyDrive/AI/Disco_Diffusion/init_images/aaa.jpg