Hey Softology, really love and appreciate what you do and I mean that with the utmost sincerity, you put in an astounding amount of work and I believe you need more appreciation. Honestly on an almost consistent day to day basis I'm still baffled that you have even a small update, it's been somewhere around a decade ago since I first used your software and back then I was using it to generate fractals! To this day I'm ignorant to how 95% of how any of this works, but nonetheless I find it fascinating and captivating. Just wanted to voice my appreciation and i'd like to pay it forward when I can! Cheers friend!
And on a side note I would like to see if I could get some insight on what could be wrong and or maybe just corrected in my case, recently my OS kicked the bucket and after a fresh installation and re-installing VOC and getting ML up and running again, scripts that had previously worked and others that worked at lower resolutions all seem to be running into VRAM issues when before they were fine. Is there something that you may know of that could be my issue or should I try starting from scratch re-installing Python, VOC, and deleting temp files to see if that possibly corrects it? Unfortunately I'm still very ignorant when it comes to reading code (might as well be Latin) so my fiddling around in the dark trying to resolve issues like these usually leads to nothing, but wasted time and I'd rather not pester you with things like this as I'm sure you're plenty busy. Thanks again!
If you get as far as VRAM memory errors then the Python etc installs should be OK. For not enough VRAM the only fixes are smaller image sizes and you can try reducing the cutout number if it is an available option. You really want a high end GPU for Text-to-Image.
@@Softology Totally understood, it was just odd to me that a lot of it was functional before reinstalling windows and subsequently VOC. Also upscaling doesn't seem to work anymore this time around, Im just gonna uninstall and start from scratch again and see if it makes a difference and if not I'll just take the loss. It's only a 3070 so in this case I get that memory is gonna be my brick wall. Thanks again!
Like all the others have said, I want to thank you wholeheartedly for making this amazing program and releasing it for free! Its so well thought out. It’s bringing us so much joy! Huge love from Bosnia ❤❤❤
Thanks a lot for this video. Today i installed vision of chaos... its blowing my mind. So awesome .... thanks !!! Too bad that i didnt ordered the 3090 instead of my 3080TI...
Is there an option to only download the models/scripts that you want to try, instead of every single one? Love the software thanks for all the hard work !
Firstly, I take my hat off you Sir for doing what you do.. Secondly, if you have a spare minute, 00:54 when you mention models that AI tools use.., who creates these models, who puts all info together for other developers of Ai tools to use ?
Have you considered adding all the models as a torrent to ease your traffic and also make it easier to download, or get more at once? I've downloaded a few multigigabyte ckpt files, each takes 2h speed dropping to few hundred kb/s, and often get hash mismatch error, so letting a torrent sync all would help.
Hey! Thank for those explanation, it is really helpful. I may be stupid but I can't find the n_batches value in VC with the Disco diffusion Turbo Script ?
I found this program today after someone on reddit mentioned it in the stable diffusion reddit. It's an amazing program but strangely I'd never seen it mentioned anywhere before. Many files have to be downloaded but that's logical. Could you add a portable mode to the installer? I changed the paths manually now but I imagine many people won't have enough space on their system drive.
Hey there, firstly love the software, i cant wait to make some crazy AI art! do you have terminal commands to kick of the generation? like say on the command line? i would like to create a few, and I don't like clicking every time.
Does this keep the machine learning models/generators updated? I'd be interested in using this for Stable Diffusion and then maybe branching out to the others, but would always like to have up to date models/weights/etc which I currently update manually whenever they're released. Also I've heard it takes 400GB to install, which models are taking up that much memory? just curious about that.
hey, thanks for this video, although I might need some further help. when I click "generate" as you do in this video, I end up with a blank image. In the status area I get the text: RuntimeError: cusolver error: CUSOLVER_STATUS_INTERNAL_ERROR, when calling `cusolverDnCreate(handle)`
Genius, thanks for your incredible work, just need to get a graphics card capable now, my GTX 680 that works ok with Mandelbrot's is no match for this for sure.
I got it to work and this software is insane even with my 3070ti. I was wondering if you knew where it looks online for reference images because some prompts will come out with a mashup of two things with similar names, like "2b2t" coming out as a mashup of 2b in minecraft
@@Softology Is it possible to use negative-weighted prompts (:-1)? If so, how? I can't get it to work in the way that I used to be able to with DD via Colab
It's a great program, but I'm asking if it's normal that every time I open this program it requires me to download something, because you say it only needs to be done once.
Hello, can I change download directory? I don't really have enough space to download all the files for machine learning stuff on my main drive, however it automatically downloads all the files in that drive. Is there a possibility?
There is now an option in version 84.2 to specify the location of the machine learning directory under the Options dialog. It will move the existing downloads to the new location too, so you do not have to re-download what you have so far.
is there anyway to speed up download times? I have 200Mb/s but it takes over 30min to download a gigabyte of data, i already tried a vpn and it didnt help
No, sorry. Download speeds are dependant on the server(s) hosting the model files which I have no control over. You only need to download them once thankfully.
Hey bro, I have the following error: RuntimeError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 4.00 GiB total capacity; 2.41 GiB already allocated; 9.74 MiB free; 2.53 GiB reserved in total by PyTorch) My PC specs are: CPU: Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz 3.60 GHz RAM: 20 GB Graphic Card: NVIDIA GeForce GTX 1050 Ti (VRAM=4 GB)
cause its giving me this error "Traceback (most recent call last): File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 1032, in loop(optim_list) File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 980, in loop train(optim_list) File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 937, in train loss.backward() File "C:\Users\adinm\AppData\Local\Programs\Python\Python37\lib\site-packages\torch\_tensor.py", line 255, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs) File "C:\Users\adinm\AppData\Local\Programs\Python\Python37\lib\site-packages\torch\autograd\__init__.py", line 149, in backward allow_unreachable=True, accumulate_grad=True) # allow_unreachable flag RuntimeError: Unable to find a valid cuDNN algorithm to run convolution Image Magick Executing command line; convert "C:\Users\adinm\Desktop\Deepdream\Project\a painting of a cabin next to a stream in a secluded forest [Zoetrope 5 ViT-B32 vqgan_imagenet_f16_1024] 307520310.png" -background black -distort SRT 0,0,1,0,0,0 -rotate 0 -distort SRT 1.02,0 -gravity center "C:\Users\adinm\Desktop\Deepdream\Project\a painting of a cabin next to a stream in a secluded forest [Zoetrope 5 ViT-B32 vqgan_imagenet_f16_1024] 307520310.png" =========================================================================== Invalid Parameter - painting =========================================================================== Finished in 068ms"
@@AdinPlays The error is "Unable to find a valid cuDNN algorithm to run convolution" which in this case indicates not enough GPU VRAM. Try a smaller image size.
you should update your help page to show the size of all the files and scripts. I installed VoC on a 2TB hard drive but for some reason all the machine learning scripts saved to my 128gb boot ssd. The entire ssd filled up so I had to uninstall python and delete the scripts lmao.
There is now an option in version 84.2 to specify the location of the machine learning directory under the Options dialog. It will move the existing downloads to the new location too, so you do not have to re-download what you have so far.
How much VRAM is recommended for this. I'm running on a 3060 with 6GB of VRAM and I simply cannot get it to start generating any image size whatsoever as it claims there's not enough VRAM available
Hey Softology, I'd really like to try this software out but Windows Defender is blocking the installation. Why is there no publisher's certificate in the install exe?
How- How do you get the images? I followed the initial tutorial but I can't figure out to import/download scripts? I keep getting blank screens when I press generate.
@@Softology So I looked down in the status and it says CUDA out of memory, and I just downloaded that earlier from one of your other videos? I'm not sure why it wouldn't be working otherwise?
That means you graphics card does not have enough memory (VRAM). Try a smaller image size. I have added more error detection to the next release so it specifically tells the user they do not have enouigh VRAM and that they should try a smaller image size.
@@AdinPlays TensorFlow is installed as part of the pip commands on the instructions page. You do need to follow all the instructions for the machine learning to work.
@@Softology i ran all the pip install commands thats why im wondering,should i reinstall the pip list or it doesnt work that way? thanks for the patience man
Ok, try the reset python steps; 1. Uninstall Python 2. Delete the directory that you installed Python into 3. Windows-R, %temp% to open your temp folder. Delete all the files within it 4. Delete the directory C:\Users\yourusername\AppData\Local\torch_extensions 5. Reboot 6. Install Python and the pip commands again 7. Reboot and then try the Machine Learning modes again That should get Python and the required libraries installed correctly.
This stuff is kickass! everything works im having one problem when rendering above image size 128 even though i have a good beefy graphics card. its all working but i get the runtimeError: CUDA out of memory. does anyone know how to allocate more memory to CUDA?
2060 has only 6 GB of VRAM. VRAM is on the GPU and what is needed for Text-to-Image. System RAM has no effect. 6 GB is minimal. A 3090 that supports up to 768x768 sized images has 24 GB VRAM.
Hey Softology, really love and appreciate what you do and I mean that with the utmost sincerity, you put in an astounding amount of work and I believe you need more appreciation. Honestly on an almost consistent day to day basis I'm still baffled that you have even a small update, it's been somewhere around a decade ago since I first used your software and back then I was using it to generate fractals! To this day I'm ignorant to how 95% of how any of this works, but nonetheless I find it fascinating and captivating. Just wanted to voice my appreciation and i'd like to pay it forward when I can!
Cheers friend!
And on a side note I would like to see if I could get some insight on what could be wrong and or maybe just corrected in my case, recently my OS kicked the bucket and after a fresh installation and re-installing VOC and getting ML up and running again, scripts that had previously worked and others that worked at lower resolutions all seem to be running into VRAM issues when before they were fine. Is there something that you may know of that could be my issue or should I try starting from scratch re-installing Python, VOC, and deleting temp files to see if that possibly corrects it? Unfortunately I'm still very ignorant when it comes to reading code (might as well be Latin) so my fiddling around in the dark trying to resolve issues like these usually leads to nothing, but wasted time and I'd rather not pester you with things like this as I'm sure you're plenty busy. Thanks again!
If you get as far as VRAM memory errors then the Python etc installs should be OK. For not enough VRAM the only fixes are smaller image sizes and you can try reducing the cutout number if it is an available option. You really want a high end GPU for Text-to-Image.
@@Softology Totally understood, it was just odd to me that a lot of it was functional before reinstalling windows and subsequently VOC. Also upscaling doesn't seem to work anymore this time around, Im just gonna uninstall and start from scratch again and see if it makes a difference and if not I'll just take the loss. It's only a 3070 so in this case I get that memory is gonna be my brick wall. Thanks again!
I found this by accident. Fate! I'm going to make some great stuff with this. Thank you so much. - Skycladsquirrelart
Thank you so much for building this!
This is awesome! If only a 1060 would work that well.
Like all the others have said, I want to thank you wholeheartedly for making this amazing program and releasing it for free! Its so well thought out. It’s bringing us so much joy! Huge love from Bosnia ❤❤❤
Thanks a lot for this video. Today i installed vision of chaos... its blowing my mind. So awesome .... thanks !!! Too bad that i didnt ordered the 3090 instead of my 3080TI...
Incredible work! I have a render rig with 4x 3090s in it. Do you know if this can utilize multiple GPUs?
I'm in love with this channel aaaaaaa
Is there an option to only download the models/scripts that you want to try, instead of every single one? Love the software thanks for all the hard work !
this is amazing!!
does it also have video input ?
Firstly, I take my hat off you Sir for doing what you do.. Secondly, if you have a spare minute, 00:54 when you mention models that AI tools use.., who creates these models, who puts all info together for other developers of Ai tools to use ?
absolut amazing thanks so much for this amazing work
I am so gratefull for your fucking amazing work! I am so inspired !
Best gui for SD.
Thank you for making this amazing tool available to the world! Will it soon support VR mode in Disco Diffusion?
It already does.
Have you considered adding all the models as a torrent to ease your traffic and also make it easier to download, or get more at once? I've downloaded a few multigigabyte ckpt files, each takes 2h speed dropping to few hundred kb/s, and often get hash mismatch error, so letting a torrent sync all would help.
Awesome work! Many thanks :)
Hey! Thank for those explanation, it is really helpful. I may be stupid but I can't find the n_batches value in VC with the Disco diffusion Turbo Script ?
Will this benefit from running on multi-GPU computers - speed wise, possible image size, or mentioned out or memory errors?
awesome program 💗
only thing I would like is image series input as well as movie as it has to extract the frames anyway
I found this program today after someone on reddit mentioned it in the stable diffusion reddit. It's an amazing program but strangely I'd never seen it mentioned anywhere before.
Many files have to be downloaded but that's logical. Could you add a portable mode to the installer? I changed the paths manually now but I imagine many people won't have enough space on their system drive.
Portable won't work. Too many system dependencies. You can move the machine learning to another drive with 1 click if your C drive is filling up.
Hey there, firstly love the software, i cant wait to make some crazy AI art!
do you have terminal commands to kick of the generation? like say on the command line?
i would like to create a few, and I don't like clicking every time.
There is no command line support.
This is amazing thanks for the tutorial
Does this keep the machine learning models/generators updated? I'd be interested in using this for Stable Diffusion and then maybe branching out to the others, but would always like to have up to date models/weights/etc which I currently update manually whenever they're released. Also I've heard it takes 400GB to install, which models are taking up that much memory? just curious about that.
Is Disco Diffusion available as a script? If so which version is supported and how often do new models get implemented?
nice clear tutorial. thanks!
hey, thanks for this video, although I might need some further help. when I click "generate" as you do in this video, I end up with a blank image. In the status area I get the text:
RuntimeError: cusolver error: CUSOLVER_STATUS_INTERNAL_ERROR, when calling `cusolverDnCreate(handle)`
What GPU and how much VRAM?
Genius, thanks for your incredible work, just need to get a graphics card capable now, my GTX 680 that works ok with Mandelbrot's is no match for this for sure.
so many steps and im stuck on the microsoft visual part because there is no /bin/hostx64. is there not an easier way i have a 3060 ti
The cuda 11.5 toolkit is available, should i stick to the 11.2.2 version ?
Dear can you make a tutorial on how to upscale after Text To Image pictures is ready? or should i upscale the Avi movie?
is there a way to utilize both graphics cards in this? or would it be worth it buying a k80?
I got it to work and this software is insane even with my 3070ti. I was wondering if you knew where it looks online for reference images because some prompts will come out with a mashup of two things with similar names, like "2b2t" coming out as a mashup of 2b in minecraft
It doesn't "look online" at all. The images come from the GAN machine learning models.
@@Softology ohh ok, i kinda assumed it used the internet to fill in any gaps in its knowledge
Hi, I was wondering are these trained on random data sets from the internet, and is there a way to train these models on your own data sets?
I'm wondering if there is a way for us to add new notebooks/methods as we find them? There seem to be new ones coming out all the time.
Send me a link and I will see if they can be added to Visions of Chaos.
Would it run on dual video card rig? Wonder if it worth getting the same one I have now and compile 12GB of RAM...
I have never tried this with a dual GPU setup.
allright when is come amd graphic card version sir ?
Never, if NVIDIA has anything to say about it. They like their monopoly and high prices.
Is it possible to use weighted prompts with Disco Diffusion 5.1 in VoC?
Yes
@@Softology Is it possible to use negative-weighted prompts (:-1)? If so, how? I can't get it to work in the way that I used to be able to with DD via Colab
It's a great program, but I'm asking if it's normal that every time I open this program it requires me to download something, because you say it only needs to be done once.
One of the servers hosting some of the models was having problems. It is now fixed. Download the latest update of Visions of Chaos and try again.
Is there anything Visions cannot do? What a silly question!
Hi, i have installed, but it doesn't show machine learning tab.
Hello, can I change download directory? I don't really have enough space to download all the files for machine learning stuff on my main drive, however it automatically downloads all the files in that drive. Is there a possibility?
There is now an option in version 84.2 to specify the location of the machine learning directory under the Options dialog. It will move the existing downloads to the new location too, so you do not have to re-download what you have so far.
Well I was hoping for high res output for print but I guess tech will have to evolve a bit xD
What happens if you ask it to draw "your mom"? I'm too lazy to download it and try
is there anyway to speed up download times? I have 200Mb/s but it takes over 30min to download a gigabyte of data, i already tried a vpn and it didnt help
No, sorry. Download speeds are dependant on the server(s) hosting the model files which I have no control over. You only need to download them once thankfully.
Do you know if it works with two Nvidia 1080 linked through SLI?
No, I don't.
Man I hope my 2070rtx 8gb will work. I just started on Google colab and I reached the data limits. Was wondering if I could run it off my computer
8GB VRAM will restrict you to images with 256x256 resolution in most of the included scripts.
@@Softology thanks now I need to find out how to resize using ai and test if it's better than just using photoshop.
I have a bunch of "Super resolution" algorithms in Visions of Chaos for upscaling.
softologyblog.wordpress.com/2021/04/23/super-resolution/
Hey bro, I have the following error:
RuntimeError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 4.00 GiB total capacity; 2.41 GiB already allocated; 9.74 MiB free; 2.53 GiB reserved in total by PyTorch)
My PC specs are:
CPU: Intel(R) Core(TM) i7-7700 CPU @ 3.60GHz 3.60 GHz
RAM: 20 GB
Graphic Card: NVIDIA GeForce GTX 1050 Ti (VRAM=4 GB)
4GB VRAM is not enough for these scripts. You really want something like a 3080 with 24 GB of VRAM.
do i use the finished processed text to image photo for the seed zoom in movie?
cause its giving me this error "Traceback (most recent call last):
File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 1032, in
loop(optim_list)
File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 980, in loop
train(optim_list)
File "C:\\Users\\adinm\\AppData\\Roaming\\Visions of Chaos\\Examples\\MachineLearning\\Text To Image\\VQGAN+CLIP\\zoetrope_5.py", line 937, in train
loss.backward()
File "C:\Users\adinm\AppData\Local\Programs\Python\Python37\lib\site-packages\torch\_tensor.py", line 255, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph, inputs=inputs)
File "C:\Users\adinm\AppData\Local\Programs\Python\Python37\lib\site-packages\torch\autograd\__init__.py", line 149, in backward
allow_unreachable=True, accumulate_grad=True) # allow_unreachable flag
RuntimeError: Unable to find a valid cuDNN algorithm to run convolution
Image Magick
Executing command line;
convert "C:\Users\adinm\Desktop\Deepdream\Project\a painting of a cabin next to a stream in a secluded forest [Zoetrope 5 ViT-B32 vqgan_imagenet_f16_1024] 307520310.png" -background black -distort SRT 0,0,1,0,0,0 -rotate 0 -distort SRT 1.02,0 -gravity center "C:\Users\adinm\Desktop\Deepdream\Project\a painting of a cabin next to a stream in a secluded forest [Zoetrope 5 ViT-B32 vqgan_imagenet_f16_1024] 307520310.png"
===========================================================================
Invalid Parameter - painting
===========================================================================
Finished in 068ms"
@@AdinPlays The error is "Unable to find a valid cuDNN algorithm to run convolution" which in this case indicates not enough GPU VRAM. Try a smaller image size.
you should update your help page to show the size of all the files and scripts. I installed VoC on a 2TB hard drive but for some reason all the machine learning scripts saved to my 128gb boot ssd. The entire ssd filled up so I had to uninstall python and delete the scripts lmao.
There is now an option in version 84.2 to specify the location of the machine learning directory under the Options dialog. It will move the existing downloads to the new location too, so you do not have to re-download what you have so far.
How much VRAM is recommended for this. I'm running on a 3060 with 6GB of VRAM and I simply cannot get it to start generating any image size whatsoever as it claims there's not enough VRAM available
You really want a 24 GB 3090. 6 GB is not enough for Text-to-Image.
Thanks for clarifying!
Hey Softology, I'd really like to try this software out but Windows Defender is blocking the installation. Why is there no publisher's certificate in the install exe?
I am not paying for a certificate at this time. If you don't trust me, don't install.
@@Softology all I needed to know! thanks for the reply. installing now...
How- How do you get the images? I followed the initial tutorial but I can't figure out to import/download scripts? I keep getting blank screens when I press generate.
You don't need to download any scripts, they are included. I haven't seen blank screens before. What does it show in the status area?
@@Softology So I looked down in the status and it says CUDA out of memory, and I just downloaded that earlier from one of your other videos? I'm not sure why it wouldn't be working otherwise?
That means you graphics card does not have enough memory (VRAM). Try a smaller image size. I have added more error detection to the next release so it specifically tells the user they do not have enouigh VRAM and that they should try a smaller image size.
I need a new video card.
On board graphics on my laptop....
Would this work on windows 11?
Yes. With a good GPU.
im getting an error saying "image magick" is not detected,and its not working,what do you suggest doing please
Install Image Magick. See softology.com.au/tutorials/tensorflow/tensorflow.htm
@@Softology now im getting en error saying that tensor flow 2 is not detected when i test and try to enable machine learning
@@AdinPlays TensorFlow is installed as part of the pip commands on the instructions page. You do need to follow all the instructions for the machine learning to work.
@@Softology i ran all the pip install commands thats why im wondering,should i reinstall the pip list or it doesnt work that way? thanks for the patience man
Ok, try the reset python steps;
1. Uninstall Python
2. Delete the directory that you installed Python into
3. Windows-R, %temp% to open your temp folder. Delete all the files within it
4. Delete the directory C:\Users\yourusername\AppData\Local\torch_extensions
5. Reboot
6. Install Python and the pip commands again
7. Reboot and then try the Machine Learning modes again
That should get Python and the required libraries installed correctly.
👍
This stuff is kickass!
everything works
im having one problem when rendering above image size 128 even though i have a good beefy graphics card.
its all working but i get the runtimeError: CUDA out of memory.
does anyone know how to allocate more memory to CUDA?
What model GPU? How much VRAM does it have?
@@Softology NVIDIA GeForce RTX 2060
Dedicated System Memory 16343 MB
this is the message i get when running in size:256 or more
also thank you for helping out
2060 has only 6 GB of VRAM. VRAM is on the GPU and what is needed for Text-to-Image. System RAM has no effect.
6 GB is minimal. A 3090 that supports up to 768x768 sized images has 24 GB VRAM.
@@Softology aaah ok thank you for clarifying
i appreciate it