ComfyUI AnimateDiff Prompt Travel: Runpod.io Cloud GPUs Tutorial

so this tutorial is going to be focused on getting run pods set up with animate diff and comfy UI it is much simpler than it seems so just follow along closely and trust me you'll get it so after you sign into runp pod and add some credits to your account you want to head over to templates here on the left and for templates you want to find runpod SD comfy UI you're going to click deploy and it's going to bring you over to this section now I'm going to use the community Cloud because it's a bit cheaper from here you can choose between various different graphics cards it's literally all your choice as long as it's available of course like you can even rent a h100 uh 4090 and you can rent previous generation as well like a a100 or a 3090 for me right now I'm just going to rent a 4,000 because it's cheaper than a 4090 so I'm going to click the deploy as you can see it has the template already set up and this start jupyter notebook so we're going to leave all that as is and we're just going to cck click continue after we click continue it's just going to give us some more specs the pricing and how much everything cost and from there we're going to click deploy so I have a previous instance that I made we're going to ignore that and we're going to focus on this new one here as you can see it's setting everything up all right so once that terminal window disappears and you see your utilization stats as well as this purple connect button it means that it is now running as you can see from the top right here so you would click connect and that's going to allow us to enter into our instance but before we do we need to set a very important thing up these three buttons here for more actions we're going to click that we're going to click edit pod and right here where it says expose TCP ports we're going to add 8188 that's the default port for comfy UI when you set it into listen mode which all make sense soon so we're just going to let it set up again so now that it's set up we're going to click connect and we are not going to connect to Port 3000 we don't want to do that instead we're going to click connect to Jupiter lab and here where it says terminal we're going to click terminal if you followed my other tutorials we are essentially in the command line but for Linux you'll notice on the left here we have a file browser system this is going to be very important to understand so if I go to the root here by clicking this folder you'll notice that there is a folder called workspace anything that is inside the workspace folder when you shut down your pod will still exist when you start your pod back up if it's outside of the workspace folder it will be deleted and that's something to keep in mind so now let's talk about getting comfy UI setup you'll notice that we have comfy UI already but we don't have a virtual environment inside our workspace folder that will allow our data to persist just in case we install some custom extensions like animate diff so the first thing we're going to do is set up a virtual environment inside of our workspace folder so we're going to python DM venv venv if you follow my other tutorials notice that it's just about the same so once that virtual environment has been created as we can see from our file browser we're going to activate it so Source vv/ bin slac activate notice that with Linux our slashes are forward slashes unlike windows and we must start with Source when we activate our virtual environment so now that our virtual environment is created we're just going to follow the standard comfy UI install steps so first thing we're going to do is CD it's the comfy UI which essentially just means that we are now inside this comfy UI folder we're going to head to the comfy UI repo and all we're going to do is scroll down to the installation instructions where you notice here it says for NVIDIA users and we are using an Nvidia card pip install everything here regarding torch so we're going to copy that we're going to head to our Jupiter lab terminal we're going to paste that in and we're going to press enter now run pods internet is for the most part pretty fast but as always you know it's 2 G 2.2 gab and it's torch so just give it a minute to install all right now that everything's finished installing the next thing that we would like to do pip install d r requirement cxt again just copy and paste go back to our Jupiter terminal paste and install all right now that that's finished installing this is a bit hard to manage with all the words on the screen so I'm just going to write clear press enter and notice it clears the screen so that we have our cursor at the top again so now we have to install the comfy UI manager to install the comfy UI manager we go to its repo we click code and copy this link here then we head back to our Jupiter lab and remember any custom nodes go inside the custom node folder so we're going to CD change directory I'm going to write custom but since I already have some of it typed out I can press Tab and it'll autocomplete the directory so now I'm inside the custom nodes folder as we can see right here and I'm just going to write get clone and paste that link the manager is now installed and we can start up our comfy UI instance so we're going to Python main.py and note-- listen all right so it's saying that our server has started and that it's located in this section so how do we navigate to this link because this link isn't going to work well we head in to run pod and as you can see if we click connect we have TCP port mappings so what we want to do is we want to take this public IP address here copy that and notice that it's telling us that the ports the internal Port would be 8188 but the external Port would be 11161 and notice that comfy UI is running at Port 8188 so we know that this external Port will be our comp UI instance so we're going to copy our link I'm going to drag this over here for clarity paste that in and then I'm going to go and copy this external Port I'm going to put that paste and notice we now have comfy UI up and running we have our manager as well and everything is is working as it should now you'll notice we have some checkpoints but they're all the standard stable diffusion checkpoints and I would imagine that that's probably not what a lot of you want to use so how do we download checkpoints shut down comfy UI I'm going to clear the screen again for clarity so to download checkpoints we're going to head into our models folder and from our models folder as you can see on our file browser there are various folders for lauras checkpoints vaes for this tutorial I'll just focus on downloading checkpoints and vaes we're going to head inside the checkpoints folder and as you can see we are now inside the checkpoints folder that was CD checkpoints and I'm going to LS LS is list list everything that's inside of this directory so you'll notice that these were all the same checkpoints that were inside of comy UI when I showed you so how do we download checkpoints in here well I'm going to go to civid Ai and find a checkpoint that I like I like this one it's really awesome I love the art style and this download button I'm going to rightclick it and I'm going to copy link address and you'll notice if I press paste and enter so you'll notice that it brings up a download menu now we're not going to download it onto our computer we're going to download it inside of our comfy UI container so how do we do that well w g w get we paste the link that we copied from civid AI again we did not copy this link up here we copied right click here and copy link address we now need to give our model a name because if we don't give it a name and an extension it's just going to name it the numbers at the end there which we do not want so we're going to dash capital O remember Capital lowercase o will result in an error so capital O and I notice the name of it here is Brick and motor mix and we're downloading version 3 so I'm just going to paste that in here B3 and safe tensors and now our model is downloaded all right it has finished downloading we're going to LS and you can see that in fact our model is now there so now before we start comfy UI back up again let's go ahead and download a vae so we're going to CD change directory and dot dot that's going to bring us out a directory we're going to CD into vae right there and we're going to W get and I have a vae here that's usually the standard stable diffusion 1.5 vae that you use the download link here on hugging face just like we did on civid AI I'm going to rightclick it copy link address and just the same if I were to paste it in here you'll notice it automatically brings in up a download head back to our Jupiter lab we have our W get command here and we're going to paste Dash capital O remember Capital lowercase will result in an error and we're going to copy the name of it right here copy and paste and just just the same you'll notice that we have our vae now we're going to CD out of this directory so now we're inside models and we're going to CD out of models into comy UI and we're going to clear screen for clarity and python main.py D- listen and again comfy UI is now up and running remember we have to head to our public IP address and our external Port which I already have set up so I'm going to head there and you'll notice that now we have our new model and if I were to load a vae in so vae loader notice that we have our vae and just to make sure everything is up and running we'll hit Q prompt and everything is indeed functional so now let's talk about getting anime different up and running because I know that's why you're all here I have my workflow file here that already has animate diff set up I'm going to drop it in boom we get our errors saying we don't have this node this node this node Etc remember that's why we installed the comfy UI manager we're going to go to manager install missing custom nodes and it gives us the options for the missing noes we're going to go one by one and install each one that one installed we click the next one that one installed we click the next one so as always to apply the installed custom nodes please restart comfy UI so we head back we contrl C and we python main.py D- listen and we get two errors the first error is associated with animate diff and the second era is that we don't have ffmpeg so all you're going to do to fix this you're going to stop the server you're going to get pull and now comfy UI is updated so if I run it again the first era is now gone but it has been replaced by a new era something that we can easily fix that is that we don't have any motion modules so how do we fix that well we're going to contrl C let's clear screen for clarity we're going to CD into custom nodes and if I go to the custom nodes folder for clarity you'll see that we have our four extensions we're going to CD into comy UI Das animate diff we LS or rather just look inside our file browser and notice that we have a models folder this is where we're going to install our models so we CD into the models folder again on civid AI we have our links I have to download I'm going to rightclick just like we did before copy link address I'm going to head in here here and just W get paste the address- capital O and name it so I'm just going to name it motion module 1 Point or5 V2 because that is the motion module that I'm downloading as you can see from here v1.5 V2 and this will be a CK PT file and we're just going to press enter so our model has finished downloading we're going to CD out of this directory CD out of that directory CD out of custom nodes and we are back at our comfy UI route which is here so if I run one last time python main.py D- listen we still have one last era and that's that ffmpeg is not installed it could not be found and this is why in the beginning of the video I'm going to shut down the server by the way I noted the importance of the workspace folder when we install ffmpeg it is not going to install inside the workspace folder so whenever you go to start up comfy UI if you shut down your pod and start it back up again which to shut down your pod you would press this stop button I'm not going to stop it just yet but when you start it back up you'll have to reinstall FFM Peg as well which it's not that hard it's really simple it's app update I'm going to clear to make things a little bit easier to see and then app install FFM Peg it's going to ask us we want to continue we're going to type Y and enter and now we're just going to python main.py D- listen never forget that D- listen so now I'm going to refresh and you'll see everything is working the way it should I'm going to click our model that we downloaded click our vae that we downloaded and I'm going to generate 16 frames as a test I'll Q prompt oh got to update our motion module here so that's the motion module that we downloaded Q promp awesome everything is working how do we download our output well in our comfy UI folder we would just output and as you can see so we would just download by right clicking and download I believe you can also download the whole folder can you download current folder as an archive and that would download everything inside of the folder all right so now the FFM Pig issue let's run into it so you guys can see what I mean I'm going to shut down and I'm going to turn off my pod why because you're going to be charged 32 cents an hour if you have a 4,000 or even more if you have a more expensive card per hour and you don't want to lose all your value so you would click stop stop pod and the entire pod is shut down I'm going to exit the Jupiter lab window because technically it's not relevant and I'm going to start up again there two things to note here one we're going to have to reinstall FFM Peg and two our Port is going to change so those are the two things that you want want to keep in mind when starting up again so when we click connect we want to head back into our Jupiter lab you'll notice that our VNV is VV is in there we're going to click terminal so we're going to Source venv SL bin SL activate we're going to CD into our comy UI folder we're going to python main.py D- listen and you'll notice everything worked except that ffmpeg could not be found so we're going to contrl C and just as before we're going to a update AP install ffmpeg hit yes so y enter and last python main.py D- listen no ffmpeg error everything's working our internal Port is the same but our external Point Port has changed so we're going to copy our link here paste it and find our external Port copy it paste it enter and we are inside of comfy UI we'll grab our workflow file again drop it set up everything run it and that is how we get anime diff set up with comy UI on runp pod make sure to shut down your pod when you're not using it because you are losing money but the great part about runp pod is that you have access to all kinds of gpus some seriously powerful gpus like 8 h100s and A1 100s as well as 490s 6000 4000s and even the older generation of 3090 TI and so on and so forth this is a great way to get up and running with anime th if you don't have compute power and this is also a great way to push your workload onto the cloud if you do have the compute power but you want to be using your computer on other tasks maybe experimenting more with comfy UI so just follow the instructions I know it may seem confusing for people who aren't the most tech savvy but after doing it a few times you'll notice that it is really easy for the most part the only commands that you're using are CD LS and W get that's really all you're doing so blessings feel free to ask questions down below I'll try to help if I can I'm considering making a Discord to make it a little bit easier to ask questions this way you know I don't have to always repeat myself but also you know we can discuss amongst each other and Al together solve bigger problems that some people are running into so blessings stay focused and stay creative

Channel: c0nsumption

Views: 7,124

Keywords:

Id: PcrwJ5C3zGM

Length: 22min 48sec (1368 seconds)

Published: Mon Oct 23 2023