Amazing video, thanks for sharing! I'd love to see a more in depth tutorial with you exploring more advanced features such as the input and target image. Can you please point me in the direction of how to apply a style transfer over existing video footage? Thanks again
HOLY SMOKES! Thank you for this. My life will now be consumed making ai videos from my Midjourney prompts. Much faster than I had anticipated. VR mode activated.
I face this error when I diffuse NameError Traceback (most recent call last) in () 68 print(f'Starting Run: {batch_name}({batchNum}) at frame {start_frame}') 69 ---> 70 if set_seed == 'random_seed': 71 random.seed() 72 seed = random.randint(0, 2**32) NameError: name 'set_seed' is not defined
I had the same issue. For me, it seemed to be due to a missing 512x512 diffusion map that didn’t download. I fixed it by switching to the 256x256 diffusion map, or by downloading one and manually dropping it into your drive
When you first connect to a runtime again, make any settings changes you want to do, enter the prompt you want to use, then go up to the runtime tab and click "run all". After your first render you can then update individual cells as you change them. I used to encounter these errors a lot, but I never got one error thrown if I run-all to get started..
Thanks for the tutorial! I've already tried this before, but learned some good tips. Can you share what you did different to create the skull face at 12:00 which appears animated somehow, but is not moving/zooming?
@@doctordiffusion Ah thank you, that's incredible! I tried doing a video init of my face and it didn't work very well...but apparently some further experimentation is in order!
Fantastic video!! How do you manage to generate that many images? (Ex. forming the video we see around 1:40?) Using DD 5.3, even with a Tesla T4 it takes several minutes to generate one image. So after 12 or 24 hours, I still feel that I will get only a very limited amount of images (even if paying). How could I manage to build a let's say 3 minutes video then? Working with less steps (ex. 120)? Working with smaller formats? Creating only several images different from each other, and use another program to "fill the space" in between? Thanks !!
I have been running Disco Diffusion locally with "Visions of Chaos" on an RTX 3080. At 512x512 I can get about a frame a minute. I will upscale and at times use After Effects or Topaz Video Enhancer for interpolation.
@@doctordiffusion so am I correct in saying that this is almost entirely unusable if you can't run it locally? I have run this with the Tesla T4 for over an hour and have 9 frames. At this rate its producing less than a second of footage an hour... For a 3 minute video this would take 3 months running non stop?
Thank you for the video. I have been creating images with disco diffusion for my other UA-cam channel. Based on your tutorial. It took me many tries to get to a prompt that works though.
What is the error code? I had an error happen where the 512x512 diffusion map wouldn’t download into my Google drive, so I had to manually download a diffusion map and drop it into the correct DD folder on drive. I also had luck switching to the 256x256 diffusion map because that one seemed to download properly. I hope we had the same error and that advice helps!!
@@rwe52496 Sorry for the slow reply but I can't even get it to run. I have a run time error. RuntimeError Traceback (most recent call last) in () 205 model.load_state_dict(torch.load(custom_path, map_location='cpu')) 206 else: --> 207 model.load_state_dict(torch.load(f'{model_path}/{diffusion_model}.pt', map_location='cpu')) 208 model.requires_grad_(False).eval().to(device) 209 for name, param in model.named_parameters(): 1 frames /usr/local/lib/python3.7/dist-packages/torch/serialization.py in __init__(self, name_or_buffer) 241 class _open_zipfile_reader(_opener): 242 def __init__(self, name_or_buffer) -> None: --> 243 super(_open_zipfile_reader, self).__init__(torch._C.PyTorchFileReader(name_or_buffer)) 244 245 RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory
@@f.a8586 try clicking "runtime" in the toolbar at the upper left of the screen. That should show you a drop-down menu where you'll see "run all". Click that and wait to see if everything runs normally. If you're connected already, disconnect from the session and then try what I said..
(urgent) i have followed every step but when i try to diffuse , i get this error runtime error : addmm_imp_cpu ' not implemented for 'half' any way to fix it ?
Mine keeps coming out without the video i put in, its jus giving me a Ai video instead of the clip i pasted on there so the Ai can actually look cool idk wat to do 😢😢
Nice video thanks. How do you chain different text prompts. I dont understand the difference between disco difusion and stable difusion. Is there a difference or different notebooks?
i get at line 218 where its do_run() " RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory" - cant get past it
Mine timed out , had to pay for more compute units, as a pay as you go and now i can't work out how to finish what i started because it time out half way through too ..how to make it shorter in future? annoying, was looking forward to see what it was going to 3d create from my prompts..
Hi is it possible to 2D animate an image we already have on our computer, for example 'zooming into it' and have a video output ? (and not and not necessarily start from a prompt)
My code is always giving these errors below in the broadcast, can you help me to solve it? NameError Traceback (most recent call last) in 51 start_frame = 0 52 batchNum = len(glob(batchFolder+"/*.txt")) ---> 53 while path.isfile(f"{batchFolder}/{batch_name}({batchNum})_settings.txt") is True or path.isfile(f"{batchFolder}/{batch_name}-{batchNum}_settings.txt") is True: 54 batchNum += 1 55 NameError: name 'path' is not defined
Great video! I have "beautiful town on a sunny day" as part of my promt and efter several attempts the background goes from sunny to green to dark brown. Anything you have noticed in your projects?
I would recommend adding “vivid colors” to the prompt but it’s a common issue that happens at times with Disco Diffusion. I believe it has to do with the area in the image that gets zoomed into not having enough detail, but that’s mostly speculation. “Hyper detailed” “in color” and “full canvas” can also help.
Try increasing the Sat_Scale. That can result in brighter colors. You can also use an init_image and then turn the Init_scale way down to retain as little of the init image as possible, so you can get something unique, while still providing the network with something it can build off of in tandem with prompts.
Great tutorial! and cool videos! I get kicked out after 1 hour even I'm using the Colab pro - May I ask how you manage to render the video after got kicked out? When I try to hit the play button at Step 5 Create the video, there always an error, the same thing happened when I tried to resume the run. Thus I only got a few frames and no video generated yet T.T
I use collab pro, and i get a captcha after 6hr sometimes, Recently i see it 20min after the run starts i click it and it can run for 12 to 18hr To make video it has to Finish 100% of the frames I always click run and then Click make video, the make video play button remains grey until the run finish And starts after the run gets to 100%. I use quick time to make A movie from frames or use Hit film
This is the way i do the run Step 1 change the text prompts and animation and settings 2d or 3d, total frames, steps, etc Step 2 go to the make video Section uncheck skip for Run all Step 3 go to the menu- run time- run all, Step 4, check at random times to click the captcha For me it is at 20 min mark Or 6hr
Damn. Mine starts off great, but then after 50 or so frames, it zooms into one color and keeps rendering out just a block of color for the rest of the run time. What do you think it is?
Every time I go to Diffuse I get the error message NotImplementedError: There were no tensor arguments to this function (e.g., you passed an empty list of Tensors), but no fallback function is registered for schema aten::_cat. This usually means that this function requires a non-empty list of Tensors, or that you (the operator writer) forgot to register a fallback function. Iv searched everywhere and cant find a fix Any Help would be amazing
Hi Doctor Diffusion. I have copied exactly what you’ve done throughout the video. I changed my prompt slightly but otherwise it’s the same. 1.1-1.6, 2, 3.. all good. Then it hits 4, Diffuse and comes up with the error, specifically a NameError: name ‘set_seed’ is not defined. I am really inspired to try get this right. I’ve been trying this for 3 days and don’t believe it’s as easy as you have made it seem in this tutorial…. Do you offer online tutoring? I really would love a human on the other side to show me how I’m messing up. I’m so inspired by artists like you and want to get on this I’m just left in perpetual frustration, tutorial after tutorial of it not working. Can you help please?
When I set my animation to 3D and run it I get an error, saying missing file dpt_large, but when I run it in 2D it works fine. Any suggestions? Thanks. Love the channel.
"Fantastic! Very good tutorial!!!. How much time can it take you to make a 6-minute video? Because the free version doesn't let me work all the time and it doesn't finish processing everything I need. And I have no idea if the paid version could help me... I need your experience. Thank you"
Thanks for turning me on to this! Question - is it “steps” under settings that determines the # of frames? I’m looking to make longer animations. Thanks!
Great video👍🏻 I just want to ask a question. Every time I try to create an animation it shows an error message saying that I’ve used up all of the allocated memory🙈 How can I fix this?
Great video bro, i have a doubt, disco diffusion makes pictures but how you add the movement?? It's with disco diffusion too? Or it's with another program?
Your video is very good explained, but by me crash every time, or it reaches up to 110 frame and then crashes and I can't make the video 😥 What can I do?
It is one of the drawback of using cloud GPU's. They can and will kick you off every now and again. You can load the last frame as a seed image and start the animation again with the same settings. This allows you to generate more frames that can be compiled together. Alternatively, if you have an RTX card you can run Disco Diffusion locally with a program called "Visions of Chaos"
@@doctordiffusion Thank you very much for your answer. Unfortunately I don't have an RTX card and I can see that they are pretty expensive. I saw that the programme offers you a monthly fee for Colab Pro, do you think that would be useful? Thanks again for your time. Have a nice day!
@@miayudanatural5029 the first tier of colab is worth the $10 in my opinion. You’ll still occasionally have things time out but it will be less frequent and you’ll gain priority on faster GPUs.
Hi, thanks so much for this really accessible tutorial! I was wondering why you take all the frames to Premiere Pro rather than use the "5. Create Video" code at the bottom of the notebook?
@@doctordiffusion ahhh I see that makes sense. I'm running my first attempt right now, excited to see the results tomorrow morning. Perhaps I'll leave a link here for whoever wants to see it.
Depends on how many frames/sec your final video be, if you create 3000 images at 24 f/s it will get you nearly 2 mins of footage. That said he mentions in the video that Topaz AI animation program can fill in by adding more in-between frames while also scaling up the images, so final animation might be a bit longer than 2 mins.
This appears to be an error affecting many Colab users including me. If I try to use Disco Diffusion after opening the Colab interface, I get the error. I can run RuDalle but not DD5.2. However, if I open Disco directly from the link in this video rather than from inside Colab, I am able to generate images. This makes little sense to me although I am glad of the work-around. Suggest you try.
@@doctordiffusion i got a Tesla P100-PCIE....and a 32GB RAM Computer. and it still would take weeks to process... what else are factors which determine the time? i am bvery frustrated as i love this stuff but will neevr get results -.-
i also had Tesla T4 and it would take 10 minutes for 1 Frame...which would be 20 days for 3000 frames :( can u help me how to make this faster? what does my computer need/requirements besides the GPU from colabs?
@@dom_wagmi lowing the amount of cutouts or your resolution will reduce times. The colab isn’t using any of your computer resources. I am dealing with the same “render times” I left it running for multiple days, feeding it the same video where it left off when it failed. I have since stopped using the colab and am doing this locally with a program called “visions of chaos” video coming soon.
@@jonhylow1239 in my Dropbox it created some folders AI - Disco_Diffusion - images_out then you should have a folder for each batch. Unsure if I will be able to assist further.
Looking for help! I've managed to create all the frames but whenever I hit the last cell to Create Video I always get an error. It's either about batchNum not being defined, or animation mode not being defined. I did set the animation mode to 3D and am not sure what is going on. Does anyone else have this or any idea what's happening? Thank you!
I have never had this part work, I always take the animation frames into an external video editor to compile. Let me know if you need some free recommendations. Sorry to hear you have had troubles. Hope this helps.
@@doctordiffusion ohh thanks so much for your reply - I thought I was doing something wrong for not getting the notebook to work haha. Any recommendations would be so helpful on how you go from where the video tutorial here ends to an actual animation ☺️ thanks so much!
@@MrMuhamad1994 the frames should all be in your google drive folder. I use after effects or Adobe premier to encode the images as a video. I am sure there are free alternatives.
@@doctordiffusion Awesome!! Now I ran into the issue of my google drive memory getting full, and it also only rendered 150 frames out of the 1000 until it disconnected. How do I make it work to render the whole thing without interruptions?
@@davidbenavides5729 it varies depending on the animation I will try out different settings until I get something I like. I’ve moved on to using Visions of Chaos and running DDv5.1 local on my RTX 3080.
@@doctordiffusion - thats awesome, and thanks for the reply. Have you noticed a speed difference running locally? I'm pretty sure we're on about the same specs based on your GPU.
as a practicing artist of 10+ years, I felt scared watching this video, with a few push of a buttons you created something that would have taken me months to create...
Knowledge and choosing is still there and prompting is a skill of a wordsmith with knowledge of specified culture. I was scared too but I'm starting to embrace it and calm down(no other way) and see it as a portable studio team for an indie developer, saving time and money
Guys I wouldn't worry at all. How many of those videos can you actually watch after getting completely bored for the lack of storytelling or a clear identity? After you watch a few you realize they all kind of look thee same. Don't get fooled by the initial "wows" of a new technology. It's just that. Another tool that artists can use, but not a replacement for the artist. Cheers!
Do you ever get the “CUDA out of memory” runtime error on the diffuse step? I get that even when I have colab pro, even following the exact steps you follow. Would appreciate any advice to resolve this!
@@zachfevans Your GPU makes no difference, check to see what GPU you're being allocated by the colab file or try reducing your image width and height.
Remember people, never support ai arts with your money or praise as they already destroying what is great about humanity which is "creations" and quality craft with thoughts and life knowledge that brought us happiness. AI art doesn't have any clue or idea for our inspiration and greatness as much as a living human itself. What corporations and ai programmers have done is not okay at all. They don't care about the quality of experience and entertainment that brought people greatness, they only care about themselves collecting huge cash and doesn't care of the earth they living. It's getting worse and worse of how they can just insult us all with these ai arts techs that will make our entertainment media worse.
It was crashing for me every now and again. Even when I was paying for colab pro. If you have an RTX card you can try “Visions of chaos” locally. It’s more stable.
Amazing video, thanks for sharing! I'd love to see a more in depth tutorial with you exploring more advanced features such as the input and target image. Can you please point me in the direction of how to apply a style transfer over existing video footage? Thanks again
HOLY SMOKES! Thank you for this. My life will now be consumed making ai videos from my Midjourney prompts. Much faster than I had anticipated. VR mode activated.
How can you incorporate midjourney into this?
This was extremely helpful! I just found another guy's video to setup. His came out after yours but it was a great intro into this!
Thanks again!
LOL I can't get anything to work. I get different errors in the code everytime I try.
you're really amazing for this
YES! This is the best tutorial I've found so far, many thanks my friend!
I face this error when I diffuse
NameError Traceback (most recent call last)
in ()
68 print(f'Starting Run: {batch_name}({batchNum}) at frame {start_frame}')
69
---> 70 if set_seed == 'random_seed':
71 random.seed()
72 seed = random.randint(0, 2**32)
NameError: name 'set_seed' is not defined
I got this same error. Did you figure out the problem?
@@tubewatcher117 Try running the "Extra Settings" Cell
@@tubewatcher117 Yes I did, some times you have to run the first task again when it says 'Reboot Required'
I had the same issue. For me, it seemed to be due to a missing 512x512 diffusion map that didn’t download. I fixed it by switching to the 256x256 diffusion map, or by downloading one and manually dropping it into your drive
When you first connect to a runtime again, make any settings changes you want to do, enter the prompt you want to use, then go up to the runtime tab and click "run all".
After your first render you can then update individual cells as you change them.
I used to encounter these errors a lot, but I never got one error thrown if I run-all to get started..
Thanks for the tutorial! I've already tried this before, but learned some good tips. Can you share what you did different to create the skull face at 12:00 which appears animated somehow, but is not moving/zooming?
That is a video of myself with white make up on my face that ran through DD. Using video source or initi images can really help.
@@doctordiffusion Ah thank you, that's incredible! I tried doing a video init of my face and it didn't work very well...but apparently some further experimentation is in order!
Fantastic video!! How do you manage to generate that many images? (Ex. forming the video we see around 1:40?) Using DD 5.3, even with a Tesla T4 it takes several minutes to generate one image. So after 12 or 24 hours, I still feel that I will get only a very limited amount of images (even if paying). How could I manage to build a let's say 3 minutes video then? Working with less steps (ex. 120)? Working with smaller formats? Creating only several images different from each other, and use another program to "fill the space" in between? Thanks !!
I second this question! Thanks for an epic vid
I have been running Disco Diffusion locally with "Visions of Chaos" on an RTX 3080. At 512x512 I can get about a frame a minute. I will upscale and at times use After Effects or Topaz Video Enhancer for interpolation.
@@doctordiffusion so am I correct in saying that this is almost entirely unusable if you can't run it locally? I have run this with the Tesla T4 for over an hour and have 9 frames. At this rate its producing less than a second of footage an hour... For a 3 minute video this would take 3 months running non stop?
@@Jack-df6ym Yep, unuseable…
Great explanation. Even as a developer, colab looks pretty menacing.
Thank you for the video. I have been creating images with disco diffusion for my other UA-cam channel. Based on your tutorial. It took me many tries to get to a prompt that works though.
My most successful promt till now is. "painting of arabian library, sunset, books, and schoolers", "trensing on artstation", "by Adonna Khare"
How do you deal with error codes? I must have deleted and reinstalled it three times, and three times I've gotten a different error code.
What is the error code? I had an error happen where the 512x512 diffusion map wouldn’t download into my Google drive, so I had to manually download a diffusion map and drop it into the correct DD folder on drive. I also had luck switching to the 256x256 diffusion map because that one seemed to download properly. I hope we had the same error and that advice helps!!
When you do your first render after connecting to a runtime, go up to the runtime tab and click "run all". I never get an error thrown if I do that..
@@rwe52496 Sorry for the slow reply but I can't even get it to run. I have a run time error.
RuntimeError Traceback (most recent call last)
in ()
205 model.load_state_dict(torch.load(custom_path, map_location='cpu'))
206 else:
--> 207 model.load_state_dict(torch.load(f'{model_path}/{diffusion_model}.pt', map_location='cpu'))
208 model.requires_grad_(False).eval().to(device)
209 for name, param in model.named_parameters():
1 frames
/usr/local/lib/python3.7/dist-packages/torch/serialization.py in __init__(self, name_or_buffer)
241 class _open_zipfile_reader(_opener):
242 def __init__(self, name_or_buffer) -> None:
--> 243 super(_open_zipfile_reader, self).__init__(torch._C.PyTorchFileReader(name_or_buffer))
244
245
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory
I've been stuck on this one forever PytorchStreamReader failed reading zip archive: failed finding central directory
@@f.a8586 try clicking "runtime" in the toolbar at the upper left of the screen. That should show you a drop-down menu where you'll see "run all". Click that and wait to see if everything runs normally. If you're connected already, disconnect from the session and then try what I said..
(urgent) i have followed every step but when i try to diffuse , i get this error runtime error : addmm_imp_cpu ' not implemented for 'half' any way to fix it ?
Congratulations, I hope I would do more videos, thanks
Mine keeps coming out without the video i put in, its jus giving me a Ai video instead of the clip i pasted on there so the Ai can actually look cool idk wat to do 😢😢
Nice video thanks. How do you chain different text prompts. I dont understand the difference between disco difusion and stable difusion. Is there a difference or different notebooks?
Thanks man you're the real G , will try it to make a music video
i get at line 218 where its do_run()
" RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory" - cant get past it
Same here!
Thank you so much. Only yours is working.
Good work Mate
Hey! I didnt understand the step from creating and image to a video. How does one chose length of video btw?
help me. is giving this error that I do?
AssertionError: Couldn't find Stable Diffusion in any of:
Mine timed out , had to pay for more compute units, as a pay as you go and now i can't work out how to finish what i started because it time out half way through too ..how to make it shorter in future? annoying, was looking forward to see what it was going to 3d create from my prompts..
For me, Thomas Kincade is a more touchy subject than ArtStation. ;P
How do you turn the 2D video to this 12:23❓ such 3d rendered depth look
I input the video sequence into blender and mixed a displacement node with a black and white version of the video.
everytime i run it i get one image and then this error:
FileNotFoundError: [Errno 2] No such file or directory: './pretrained/AdaBins_nyu.pt'
hi, thanks for the vid ! which version of colab are you using ?
How can you continue learning? Google Colab is cut off in the middle due to runtime restrictions.
You can start a new session use the same prompts and settings. Be sure to use the last rendered frame as the starting image seed.
@@doctordiffusion Where do I enter the last rendered frame?
Damn dude that is insane
thank you ❤❤
Thank you for sharing this walkthrough. Can you close the tap/computer while it runs, or do you have to be online throughout?
It will likely disconnect you after a bit if you do not have the page open.
Hi is it possible to 2D animate an image we already have on our computer, for example 'zooming into it' and have a video output ?
(and not and not necessarily start from a prompt)
Okay i need help, anyone here know how to fix the "set_seed" is not defined
My code is always giving these errors below in the broadcast, can you help me to solve it?
NameError Traceback (most recent call last)
in
51 start_frame = 0
52 batchNum = len(glob(batchFolder+"/*.txt"))
---> 53 while path.isfile(f"{batchFolder}/{batch_name}({batchNum})_settings.txt") is True or path.isfile(f"{batchFolder}/{batch_name}-{batchNum}_settings.txt") is True:
54 batchNum += 1
55
NameError: name 'path' is not defined
How do you change the length of the video. I want a 10 seond video for example so what settings would i adjust for that?
also what about FPS?
Thanks for this tutorial ! Could you share your Topaz AI settings please?
thank you
Sir, what music plays in background?
How long would you say it took to create the rainbow lighthouse frames?
Ok, I can do this,thx for tut
Great video! I have "beautiful town on a sunny day" as part of my promt and efter several attempts the background goes from sunny to green to dark brown. Anything you have noticed in your projects?
I would recommend adding “vivid colors” to the prompt but it’s a common issue that happens at times with Disco Diffusion. I believe it has to do with the area in the image that gets zoomed into not having enough detail, but that’s mostly speculation. “Hyper detailed” “in color” and “full canvas” can also help.
@@doctordiffusion Thank you!
Try increasing the Sat_Scale. That can result in brighter colors. You can also use an init_image and then turn the Init_scale way down to retain as little of the init image as possible, so you can get something unique, while still providing the network with something it can build off of in tandem with prompts.
Thanks !
How did you made the smiling Joker? looks awesome! Thank you for lovely video. Answers most of my questions.
Can we make a one hour animation with this ?
What is the name of the last song at the end of the video ?
Thanks !
Great tutorial! and cool videos! I get kicked out after 1 hour even I'm using the Colab pro - May I ask how you manage to render the video after got kicked out? When I try to hit the play button at Step 5 Create the video, there always an error, the same thing happened when I tried to resume the run. Thus I only got a few frames and no video generated yet T.T
I use collab pro, and i get a captcha after 6hr sometimes,
Recently i see it 20min after the run starts i click it and it can run for 12 to 18hr
To make video it has to
Finish 100% of the frames
I always click run and then
Click make video, the make video play button remains grey until the run finish
And starts after the run gets to 100%.
I use quick time to make
A movie from frames or use
Hit film
This is the way i do the run
Step 1 change the text prompts and animation and settings 2d or 3d, total frames, steps, etc
Step 2 go to the make video
Section uncheck skip for
Run all
Step 3 go to the menu- run time- run all,
Step 4, check at random times to click the captcha
For me it is at 20 min mark
Or 6hr
@@SHAGGYMALO how long does it take for you to make it all frames...after 1 hour i have still 8/10 000 frames....will never finish
@@dom_wagmi dude i have the same question lol taking literally forever what the heck xD
Amazing
Damn. Mine starts off great, but then after 50 or so frames, it zooms into one color and keeps rendering out just a block of color for the rest of the run time.
What do you think it is?
Is it possible to show how you actually make it into a video?
Every time I go to Diffuse I get the error message NotImplementedError: There were no tensor arguments to this function (e.g., you passed an empty list of Tensors), but no fallback function is registered for schema aten::_cat. This usually means that this function requires a non-empty list of Tensors, or that you (the operator writer) forgot to register a fallback function.
Iv searched everywhere and cant find a fix
Any Help would be amazing
HI.. I need a tutoria to run locally! can you help me?
Hi Doctor Diffusion. I have copied exactly what you’ve done throughout the video. I changed my prompt slightly but otherwise it’s the same. 1.1-1.6, 2, 3.. all good. Then it hits 4, Diffuse and comes up with the error, specifically a NameError: name ‘set_seed’ is not defined. I am really inspired to try get this right. I’ve been trying this for 3 days and don’t believe it’s as easy as you have made it seem in this tutorial…. Do you offer online tutoring? I really would love a human on the other side to show me how I’m messing up. I’m so inspired by artists like you and want to get on this I’m just left in perpetual frustration, tutorial after tutorial of it not working. Can you help please?
When I set my animation to 3D and run it I get an error, saying missing file dpt_large, but when I run it in 2D it works fine. Any suggestions? Thanks. Love the channel.
Hmm, not really sure. I’ll update if I find any information or a solution.
"Fantastic! Very good tutorial!!!. How much time can it take you to make a 6-minute video? Because the free version doesn't let me work all the time and it doesn't finish processing everything I need. And I have no idea if the paid version could help me... I need your experience. Thank you"
Thanks for turning me on to this! Question - is it “steps” under settings that determines the # of frames? I’m looking to make longer animations. Thanks!
How do I make it run longer with lower steps ? I did 500 but it only comes out with 12 pictures
do you mind sharing a bit on the 2d animation? how to set the x and y movement?
how do i use my own images i made and not the text input?
Mine is just stuck at 0%, this makes no sense to me
Great video👍🏻 I just want to ask a question. Every time I try to create an animation it shows an error message saying that I’ve used up all of the allocated memory🙈 How can I fix this?
Can you show us how you used Topaz video enhance AI to upscale slow and add in between frames please?
Great video bro, i have a doubt, disco diffusion makes pictures but how you add the movement?? It's with disco diffusion too? Or it's with another program?
He does explain near the end of the video how he creates the animation and which programs he uses.
Hi! thanks for this! We can do it with the free version of Colab?
Your video is very good explained, but by me crash every time, or it reaches up to 110 frame and then crashes and I can't make the video 😥 What can I do?
I have a new PC with 64bit
It is one of the drawback of using cloud GPU's. They can and will kick you off every now and again. You can load the last frame as a seed image and start the animation again with the same settings. This allows you to generate more frames that can be compiled together. Alternatively, if you have an RTX card you can run Disco Diffusion locally with a program called "Visions of Chaos"
@@doctordiffusion Thank you very much for your answer. Unfortunately I don't have an RTX card and I can see that they are pretty expensive.
I saw that the programme offers you a monthly fee for Colab Pro, do you think that would be useful?
Thanks again for your time. Have a nice day!
@@miayudanatural5029 the first tier of colab is worth the $10 in my opinion. You’ll still occasionally have things time out but it will be less frequent and you’ll gain priority on faster GPUs.
Tumultuous - TUH MUL CHU US
Hi, thanks so much for this really accessible tutorial! I was wondering why you take all the frames to Premiere Pro rather than use the "5. Create Video" code at the bottom of the notebook?
I normally run until it crashes rather than have it stop after an amount of frames. I also like the amount of control I have in other programs.
@@doctordiffusion ahhh I see that makes sense. I'm running my first attempt right now, excited to see the results tomorrow morning. Perhaps I'll leave a link here for whoever wants to see it.
@@sephwilson5127 Interested in seeing the results!
@@sephwilson5127 we are waiting for link)
@@sephwilson5127 would love to see this!
Is colab pro required?
Thanks for the guide! any idea why disco is generating random frames that doesn't connect? just new unique image for every frame
Did you make sure to select an “animation type”? :)
where can i set up how many frames per seconds it does? or whats default? if i choose 3000 frames, how long will the final video be?
Depends on how many frames/sec your final video be, if you create 3000 images at 24 f/s it will get you nearly 2 mins of footage.
That said he mentions in the video that Topaz AI animation program can fill in by adding more in-between frames while also scaling up the images, so final animation might be a bit longer than 2 mins.
Having an error to make video using frame
On number #2 it seems stuck on this error "first URL failed using fallback "
This appears to be an error affecting many Colab users including me. If I try to use Disco Diffusion after opening the Colab interface, I get the error. I can run RuDalle but not DD5.2. However, if I open Disco directly from the link in this video rather than from inside Colab, I am able to generate images. This makes little sense to me although I am glad of the work-around. Suggest you try.
What batch number error can be sort out
The diffuse keeps failing what did i do
how long does this process take for you? after 45 minutes i am still 8/10000 frames -.- this will never end
It varies depending on what GPU you are assigned in the colab instance.
@@doctordiffusion i got a Tesla P100-PCIE....and a 32GB RAM Computer. and it still would take weeks to process... what else are factors which determine the time? i am bvery frustrated as i love this stuff but will neevr get results -.-
i also had Tesla T4 and it would take 10 minutes for 1 Frame...which would be 20 days for 3000 frames :( can u help me how to make this faster? what does my computer need/requirements besides the GPU from colabs?
@@dom_wagmi lowing the amount of cutouts or your resolution will reduce times. The colab isn’t using any of your computer resources. I am dealing with the same “render times” I left it running for multiple days, feeding it the same video where it left off when it failed. I have since stopped using the colab and am doing this locally with a program called “visions of chaos” video coming soon.
@@doctordiffusion this program looks insane! is there any chance to use this with rented GPU? my pc is not very good :/
I feel like there's stuff I need to learn before I start with this tutorial but I don't know where. Any suggestions? Thanks.
😱😱😱
Can Disco Diffusion be used to generate music visualizations?
of course
look at people's music videos
Hey, when I make 5 batches, how to save the first 4 ? I only can see the last one of them. Can you help me ?
They should be in your google drive folder :)
@@doctordiffusion only the last one appears to be in there, unless Im not completely blind. Can that be ?
@@jonhylow1239 in my Dropbox it created some folders AI - Disco_Diffusion - images_out then you should have a folder for each batch. Unsure if I will be able to assist further.
@@doctordiffusion ok thanks for trying to help. Maybe Im gonna find it now.
@@doctordiffusion I use google drive and there arent any visible folders.
crashed all the time for me.... :(
Looking for help! I've managed to create all the frames but whenever I hit the last cell to Create Video I always get an error. It's either about batchNum not being defined, or animation mode not being defined. I did set the animation mode to 3D and am not sure what is going on. Does anyone else have this or any idea what's happening? Thank you!
I have never had this part work, I always take the animation frames into an external video editor to compile. Let me know if you need some free recommendations. Sorry to hear you have had troubles. Hope this helps.
@@doctordiffusion ohh thanks so much for your reply - I thought I was doing something wrong for not getting the notebook to work haha. Any recommendations would be so helpful on how you go from where the video tutorial here ends to an actual animation ☺️ thanks so much!
@@doctordiffusion okay cool where and how do you do that part for the animation?
@@MrMuhamad1994 the frames should all be in your google drive folder. I use after effects or Adobe premier to encode the images as a video. I am sure there are free alternatives.
@@doctordiffusion Awesome!! Now I ran into the issue of my google drive memory getting full, and it also only rendered 150 frames out of the 1000 until it disconnected. How do I make it work to render the whole thing without interruptions?
How many steps are you doing?
Nice Tutorial,
You should show step 5 as the other guys since I always get an error there too
I never have used step 5. I take the images rendered out in my google drive folder and then import them as an image sequence in Adobe Premiere.
@@doctordiffusion that’s a great method!
@@doctordiffusion - thanks for the info here, what settings do you use for Topaz Video Enhance AI? Anything you can share please? Thanks!
@@davidbenavides5729 it varies depending on the animation I will try out different settings until I get something I like. I’ve moved on to using Visions of Chaos and running DDv5.1 local on my RTX 3080.
@@doctordiffusion - thats awesome, and thanks for the reply. Have you noticed a speed difference running locally? I'm pretty sure we're on about the same specs based on your GPU.
as a practicing artist of 10+ years, I felt scared watching this video, with a few push of a buttons you created something that would have taken me months to create...
I agree. We could be screwed. Making a video about that
Knowledge and choosing is still there and prompting is a skill of a wordsmith with knowledge of specified culture. I was scared too but I'm starting to embrace it and calm down(no other way) and see it as a portable studio team for an indie developer, saving time and money
Guys I wouldn't worry at all. How many of those videos can you actually watch after getting completely bored for the lack of storytelling or a clear identity? After you watch a few you realize they all kind of look thee same. Don't get fooled by the initial "wows" of a new technology. It's just that. Another tool that artists can use, but not a replacement for the artist. Cheers!
Too-mulch-you-us
As of now is 5.7 xD
is this video still relevant
Do you ever get the “CUDA out of memory” runtime error on the diffuse step? I get that even when I have colab pro, even following the exact steps you follow. Would appreciate any advice to resolve this!
Whats your graphics card?
@@SauceBlob I have a 3090 and still get this. hmm
@@SauceBlob Your GPU doesn't make a difference when you're running a program off someone elses GPU.
@@zachfevans Your GPU makes no difference, check to see what GPU you're being allocated by the colab file or try reducing your image width and height.
I'm getting RuntimeError: CUDA error: an illegal memory access was encountered. I can never get any of these colabs to work
im only getting an image
You will only get one image at a time but it should continue to save a sequence of images that can be used to create animations in other software.
@@doctordiffusion im only getting 1 image
@@AnimeGamerCreater Did you expand the "animation settings" tab and select an animation mode? 3D is the one I chose for this video.
@@doctordiffusion yea
@@AnimeGamerCreater what happens when the first image finishes? The colab stop?
Hello bruh is that uh?
Remember people, never support ai arts with your money or praise as they already destroying what is great about humanity which is "creations" and quality craft with thoughts and life knowledge that brought us happiness. AI art doesn't have any clue or idea for our inspiration and greatness as much as a living human itself.
What corporations and ai programmers have done is not okay at all. They don't care about the quality of experience and entertainment that brought people greatness, they only care about themselves collecting huge cash and doesn't care of the earth they living. It's getting worse and worse of how they can just insult us all with these ai arts techs that will make our entertainment media worse.
Google Collab is trash. There must be a better way.
u kinda remind me of @allmyfriendsaresynths :D
aaand a hundred years later you get a 5 second video lol
After running step 4 diffuse, my session crashes for an unknown reason. Is there a way to fix?
It was crashing for me every now and again. Even when I was paying for colab pro. If you have an RTX card you can try “Visions of chaos” locally. It’s more stable.