Disco Diffusion for ABSOLUTE BEGINNERS (Make your 1st image TODAY, for free!) - Tutorial Part 1
Вставка
- Опубліковано 24 сер 2024
- Disco Diffusion V5.2:
colab.research...
This is a more simple and more precise Tutorial about Disco Diffusion. If you like it, your can watch this one: • Quick & Easy Tutorial ... - you'll learn other things (about steps, for example).
Yes, I added a sort of Christmas music for this one, because to be able to build these images is like Christmas, right?
/ quick_eyed_sky (to support, get prompts, ideas, and images)? I have 79 Patrons today: thank you!!
You can also send me a little money with the new "Super Thanks" button under the clip. Merci !
Playlist of all my AI tutorials: • Tutorials for Disco Di...
If the colab crashes with a lack of memory, just restart it and ask for a "fewer pixels" picture, like 512 x 384. If it doesn't work, disconnect and restart everything. Clean your Google Drive.
You can also try other Disco Diffusion colabs:
colab.research...
colab.research...
There's a Discord you should enter:
/ discord
Many Colabs are described here: softologyblog.... - try them!
One Disco Diffusion FAQ: docs.google.co...
Have fun!
Liked it? / quick_eyed_sky if you want to help and say thanks.
Here are the multiples of 64:
64
128
192
256
320
384
448
512
576
640
704
768
832
896
960
1024
1088
1152
1216
1280
1344
1408
1472
1536
1600
1664
1728
1792
1856
1920
1984
2048
2112
2176
2240
2304
2368
2432
2496
2560
2624
2688
2752
2816
2880
2944
This was the most useful tutorial I've found so far. Thank you for clearing up so many things.
You are one Based ambassador of AI
Thank you for this tutorial. You saved a lot of trouble trying to figure this out on my own. Even saved me some trouble by providing usable multiples of 64! Thank you!
Never seen a tutorial so clear and precise. I tip my hat to you.
Merci François !
@@QuickEyedSky Merci à toi !
thank you so much! the best way to enter into the AI art is this tutorial!
omg, i finally made my first one!! thank you so much
Finally. Man you are the best. They made a good AI but they didn't make a human interface for it so its totally impossible to use it till smb explains. You did it the best way. Thank you.
This one is for me, an absolute beginner! Thank you! Cute music!
Excellent resource, thanks Sky!
Great video thanks
Thank you so much this video made so much more sense than the other tutorials i watched
Thank you for sharing, you bring me to the dreams
thank you so much for the tutorial! very clear, neat and to the point. Thanks again! And with beautiful examples :)
I was like you, you know, a bit frightened by all the code and settings. I had to make a tutorial because I couldn't find any. This one is a more precise one. Thanks again!
Awesomeness!
Nice easy tutorial! I'm sure this will help a lot of beginners! 👍
Fantastic intro, thank you once again.
You are a true true hero
thx for that - very helpful!
Thank you for this amazing tutorial.
I really wanted to make one of these! but google warned me that it will required more ram and process will be canceled. I hope someday when I get a good pc I will make these beautiful images
The hero we needed!
Thanks
nice tutorial tyvm
Amazing tutorial. Thank you for sharing - I enjoy your enthusiasm as well. =]
Wow thanks soo much
I love this but my computer doesnt complete de process, it shows a red bar and a red graphic. Don´t know if someone ran into this issue. Help please?
Superb, many thanks.
Can I use this for commercial use?
Yep
Excellent tutorial . . . which leads to another question. What if I want to create something truly abstract, more than a bit different than what seems to be the common theme with most of the DD images i've seen so far? OK, I'm missing something here, but what sort of image component types can be processed into this, beyond buildings, clouds, robots, etc. . . . how about completely abstract fractal constructs, biologicals, and so on?
If there is an answer for that, amazing. Of course, I may have completely missed how this actually operates, but that's the concept . . .
Yes, yes!! Just ask for it in the prompt: "abstract fractal", add words like shaded or cubism. Etching or "intricate velvet molecular happiness". You can ask everything...
Thanks so much!, great tutorials!
Hi Quick Eyed,do you know if this programmed could help me tansform a drawing made by hand to digital?
Thank you so much ❤️
Do you bring your images into photoshop or another painting program to alter them? Thanks for the great tutorials!
what is stitching ? can you make a video explaining how to do it
Valeu!
I get an error on “diffuse!”:
CUDA out of memory. Tried to allocate 10.83 GiB (GPU 0; 14.76 GiB total capacity; 8.10 GiB already allocated; 5.71 GiB free; 8.25 GiB reserved by PyTorch) If reserved memory is >> allocated memory Try setting max_split_size_mb to avoid fragmentation.
What should i do??
Thanks ! I currently use nightcafe studio, I was wondering except that you don't have limited credits with disco diffusion, what are the differences between this and using nightcafe's coherent setting ?
I wish all the text to image a I were as straight forward as dall e mini but thanks anyway cos I didn't stand a chance without this video. How did they expect newbies to figure this out themselves??? 🤷 I'd have to watch a couple more times.
That's why I made these videos. Python code and settings are frightening for the common human non-geek being.
Mine always stops at "1.5 Define necessary functions" how can I fix this?
I follow this and it is telling me: if steps
mines too
Why is a lower dispaly rate better ?
wow,
thanks
One question, how long does it take to be able to use DD again using the free version? After a few images it tells me that I have reached the limit! Excellent video, thank you so much!
Google is smart, they want you to pay for Pro. Frankly, I don't remember. A few hours, I suppose. If you have pleasure exploring this tool, I suggest you take Pro for a month, to see. It gets faster (25 min/image), for like 12 hours before it stops. Of course, at this stage, after a month I took Pro+ (7 minutes/images, 2 colabs with V100s, 24/24).
@@QuickEyedSky interesting! i'm contemplating getting pro, and maybe eventually pro+. is it reliably faster in your case? i've seen many people complain that they pay for pro+ but only get a marginal increase in speed.
@@QuickEyedSky Thank you so much for your answer! Keep making this amazing work!
I dont see where I can hide the code at 1:43. Where do I have to click? I dont find it.
You just right-click into the code area.
thank you !
and i love the music 🤩
This doesn't work for me. When I try to hit the "start" buttons for all of the run settings I get errors. In basic settings I get "name 'diffusion_model' is not defined" and then in Extra Settings, I get "name 'batchFolder' is not defined." Then I can't do Do The Run because I get "name 'animation_mode' is not defined." This shit is so counterintuitive...
I'm an absolute noob at this and keep getting the same error prompt : NameError Traceback (most recent call last)
in
4 n_batches = 50 #@param{type: 'number'}
5
----> 6 if animation_mode == 'Video Input':
7 steps = video_init_steps
8
NameError: name 'animation_mode' is not defined
Can somebody help me with this please ? thank you
What is the minimum RAM needed to create animations using this tool? for now, I only have a MacBook Pro with 8GB RAM
It happens online, not on your Mac. This is the principle of colab.
@@QuickEyedSky is there a reason it’s so slow? It took 3 hours to get 5 images
wtf, i've got error, but my setting is the same as you... now i get a headache!
So.... The results that we get is just randomly images right? Or we can decide what images we can make?
Please answer it..
I kinda confused about the result of image came from..
The AI tries to follow your prompt. And yes, you'll be surprised!
@@QuickEyedSky aah oke, what if i enter 10 on intermediate_saves? Is that make the image more detail?
Got runtime error?
If there a way to run multi initial images and multi prompts together?
You can watch this tutorial and think about cells: ua-cam.com/video/-uQ6x-kkjMk/v-deo.html
What's the best vertical ratio for phone?
It depends on the phone... Google is your friend, here.
Is it ok to close browser when it's running?
No, only if you have Pro+
Does the GPU that you get affect the pictures or only the speed
The speed only. It's usually one hour on free, and 25 mn, or 7 minutes on Pro.
muito bom seu tutorial, estou ancioso pra começar minhas imagens. o que seria esse erro?
File "", line 5
image_prompts = {
^
SyntaxError: invalid syntax
Em image_prompts adiciona a sua própria imagem e a ai irá recriá-la.
*Parece que não forneceu uma imagen, mas não tem de o fazer.*
@@LuigiCotocea Obrigado!
How much gpu do you need generally?
On Pro+, I work 2 colabs at the same time. Both with V100s, sometimes a V100 and a P100.
@@QuickEyedSky what a time to be alive with this. Hard to imagine where this will end up
@@delatroy I think it'll be bigger than we imagine. Millions of people will begin to create images.
@@QuickEyedSky I feel like its a double edged sword. No one will understand what's real and invented anymore so governments can use it to invent reality.
Hi friend, following this tutorial on a basic free plan I'm getting this error message when at Diffuse step. "RuntimeError: CUDA error: an illegal memory access was encountered" It managed to make 18 steps out of 240 and then it bailed on me with that error. Trying to re-run step 4 it gives the same error multiple times. Is this something that is usual on a free plan? Will it go away on paid plans? I'm trying to render low-res pic as you suggestged with 5 dispay_rate and 2 batches... Looking forward to your feedback :)
Try tomorrow. Try Disco Diffusion 4 too. You'll make on image, you'll see.
You really need to pay for the $10 a month pro version of Google Colab. You can easily get 30+ large images a day. That's more than enough for the image sizes I want to create (1280 by 768). Each of mine takes about 30 minutes each. That's about 10 to 15 hours of time. Of course you can reduce the size of your images and their quality to get more images faster. But I prefer the quality over the quantity.
@@henrythegreatamerican8136 Thanks for this. It's a good choice/balance. I have Pro+, I often get 2 V100. An image every 7 minutes, but it's $50 a month.
Man, this program seems super slow
lol you're not showing it easy but at least you're having fun editing your videos
Bro you have any social media where I can contact you....I want to create a PERFECT ART in Disco Diffusion but I just couldn't 😔 Maybe I should word my prompts better