Warp Fusion: Step by Step Tutorial
Вставка
- Опубліковано 2 лип 2024
- Warp fusion is a fantastic AI animation tool to create videos that just pop. In this video I show you step by step how to use warp fusion using a remote GPU. This is my preferable method as it allows me to run multiple GPU simultaneously freeing up my PC to work on other projects.
I have been playing with Warp fusion for weeks and honestly had so much fun with it, check out my twitter and Instagram for my results.
Links for everything I use in this video:
🛠️ / posts
🛠️ Google colab: colab.research.google.com/
🛠️ Model: civitai.com/models/4823/delib...
🛠️ Lora: civitai.com/models/83872/ener... website: 🛠️www.promptmuse.com
Help & Support:
/ discord
♥️ My settings file is here for you to download:
Stone colored statue lady :
👉drive.google.com/file/d/1cVsV...
White colored statue lady:
👉drive.google.com/file/d/1v1_h...
Timestamps:
Coming soon
🔊 Socials, come and say hi:
Instagram: / promptmuse
Twitter: / promptmuse
Developer Credits:
Massive thanks to Alex the developer of Warp fusion for his ongoing dev and support. Also, huge shout out to Kytra (Those on discord will of probably received some support from Kytra at some time or another).
FallenIncursio for creating the energy vein Lora and XpucT for the fantastic Deliberate model. - Навчання та стиль
I will endeavor to set a FAQ here. But, the discord server for warp fusion is very helpful, if you are feeling a bit lost: discord.gg/vXBTtzmeJd
Not feeling too lost, your tutorial was one of the most understandable of all the controlnet videos! Please keep up the great work! :)
This is the best one, whattt, it's so straightforward, and I LOVE that it's literally all the next relevant information. No anecdotes, or any of that, just a straight up guide. THANK YOU
Your tutorial and explanation is by far the best. Everything is detailed in such way that one can actually understand what is for what. Others who have made tutorials on same, honestly, crap big time. Thank you thank you thank you
Your attention to detail is second to none. I'll be using this method on my next music video. I can't thank you enough Queeny x
Thank you so much beautiful ❤ Goodluck with you music video (I’d love to see it when it’s complete) x
You're probably getting bored by all the kudos, but I have to say that your tutorial has helped me A LOT!!! Thank you!!!
I was already a Patreon for WF but hadn't got my head around it. Brilliant tutorial and thank you!
Awesome, get some use out of your Patreon 🤣 👍 You are very much welcome !
I followed along and am doing the run now tweaking my prompts! Was not able to get this far w any other tutorial, Thank you so much!!!! Subbed.
You are so welcome!
Thank you for the detailed explanation. Now I’ll finally use Lora models after putting them off for so long 😅
Another great tutorial for something I thought I had no interest in and now I want to try!
Awesome, get those creative juices going! if you get stuck any point their discord is great place to ask questions. It can e a bit confusing at the start. (Link is in the description) :)
Very Nice, cant wait to try!
Nice tutorial! Thumbs up for you. You really helped us.
Great, i am happy you got to the warpfusion, it is really amazing algorithm and rendering 1920x1080 is not an issue anymore. ✌Thanks for tutorial. Also I found out that adding temporalnet is giving really better result than deflicker in Davinci. Although I am applying both.
Totally 👌 I think deflicker will be a none issue soon with CN and scripts. It’s frustrating as I prefer AE over davinci. Looking forward to perhaps not needing that extra step soon.
This was extraordinary helpful ❤
Fantastic ! Thank you for letting me know, glad it was helpful 🔥🫶
Ive been trying to get this working but thanks to you i managed to pass through the errors, cheeeers!
The Deflicker effect is an OFX effect that can be use directly from the Edit page.
It saves a small step by not using Fusion.
Amazing, clear tutorial as always. Thank you so much!!
You're very welcome! I’m glad you liked it!!
Amazing tutorial!!!
Thank you so much, hoping to do a deeper dive video soon!
Thanks Prompt Muse. I think I'll have a go at making one of these. PS. Nice house 😁
Awesome, did you say in the chat you are running locally?
nice tutorial
Nice video ! 👍
Thank you !!
Probably the clearest Tutorial on Warp Diffusion out there! Nevertheless, it is not an easy one to make, there are just so many steps compared to other AI's...
I will give it a go and see If I make it😅
Thank you so much! All I can say is take your time and have fun. 🔥🦾
Thank you so much! All I can say is take your time and have fun. 🔥🦾
Relatively new to this and omg… 😱 this is literally A LOT to take it. Will likely rewatch this. Curious to know if there’s an online course that goes in depth. Would love to learn this for work.
enjoyed your sharing, really is quite a lot of grasp but nevertheless it was very clear and I love it
Glad you enjoyed it!
I have cool AI images and would love to animate them. Cool! cant wait😊
Awesome, so many ways now. Some are easier than others. This process may be a little bit more work but produces great results. There is also deforum and batch images in stable diffusion. Everything produces somthing slightly different which is very cool.
yipeeee. cheers
🫶🦾 You are welcome
HELLO! This is like literally the first time I've come across ur channel and truth be told, you know those people you judge whom you assume that just because they're good-looking meant naturally that it makes up for other things; one of them being intelligence and then when it turns out otherwise you secretly slap yourself and reflect on how judgemental you were?
Well, it just happened to me. Your video was so clear and informative it guided me through my process of making my first warpfusion video so smoothly and the reason why I chanced upon yours was because the first two I tried either skipped important info during the process (maybe because to them it seems unneccessary and I'm a total noob or what idk) but it frustrated me to the point I went searching for other tutorials and landed with yours! So thank you! I just subbed to your page! (Once again, sorry for being the presumptous prick...I'm only human:( Looking forward to more videos!)
@promptmuse do spelling errors affect your prompt results? Cause I saw your first line of prompt written (as quoted):
"text_prompts": {
"0": [
"a brauitful woman as a (stone_color,stone_statue:1.3), "
Might wanna consider editing it for future learners!
haha Sometimes, but generally it can recognise the word. If the syntax is wrong then it will not work but my terrible grammar can be passible in most cases.
Love the accent so proper
I like your eyes, and thanks for the video!
Great video there. I have 2 questions. what do you specify for custom_embed_dir under lora path. Also if you want to transform a video with more than one character, is it possible to use multiple Loras in one transformation. Thanks
Thanks!
Whatttt amazing dude, Thank you 🫶🫶🫶🔥🔥🔥
@@promptmuse You make the best tutorials - I really appreciate it!
Brill tutorial. You've explained the GUI settings much better than any other tutorial. Would you know of the best free application for deflickering animation? Lots of the AE plugins or Divinci resolve, are very expensive.
Short answer is davinci in my opinion has the best deflicker (But yeah, paying for just the use of that plug in sucks 😩) got a couple of codes on github I’m going to test out next, and post the results, so please hang around to see. I’ve tested quite a few and they all were meh. But, I have faith a decent one will be along for an eighth of the price if not free, soon. Especially with Control nets being developed so fast and furiously!
@@promptmuse thanks for your reply. Love all your tutorials.
Hello ! Great video ! I just don't understand what controlnet is ? When you fill the controlnet_models_dir :
What's the difference between this and the setting.txt file ?
Great video. Thanks for the detailed tutorial! For doing variations of the same init video, what do you recommend to do? For example, I just want to change the "model", I load the new model on my settings folder but when I run the "Run before", all cells crash...seems i need to do it all over again from scratch... is there any other way? Thanks!
Hey all you need to do is change the model path. Then rerun that cell (leave all the others) and then run diffuse… bobs your uncle…. or monkey or summit like that.
No need to rerun every cell 👍
thanks for this! but what is the purpose of a checkpoint? I can make my own checkpoint?
You ever figure out how to add extensions in there? like Adetailer,etc
I cannot find this notebook on the patreon page and your google notebook link is not to this notebook but to the general google colab page. please link this version of the notebook you are using. thanks
Thank you for the tutorial, what is this line in your setting text
/content/drive/MyDrive/Misc/control_v11p_sd15_inpaint.pth
Very professional presenting, you belong on prime time television. reminds me of tomorrows world on the bbc 👍.
Thank you ! I use to absolutely love Click and the gadget show 😂 Grew up on these !😊
Not sure what I'm doing wrong, but every video I'm trying to render is showing with the prompt 'Frame 1 Prompt: ['illustration of a robot with futuristic glasses, ect ect'. I've tried changing the default prompt multiple times, but it always seems to run it with that same prompt. Any idea on why this is happening?
Honestly, the Turtorial was what im looking for, but the Beautiful Women i saw was more insteresting xD haha =P
why change "load_to:" setting to "GPU"...been running google collab with the CPU option and its working...Any incite on this?
Isn't deflicker available only in paid version of Davinci? Just curious cause last time i check it was only in Studio version :(
Sadly yes. I actually prefer AE, there is a deflicker plug in that you can buy. But still costs quite a chunk. I’m looking at free deflickers today, but will only publish if they are any good.
I think there will be a free deflicker built into control net or a github version soon. Someone will be working on it.
Hi, my big problem is render time (collab pro), it takes 2 hours for 30 sec, 720p render. Any tips to speed up?
Oooouuuff, Increase the extract nth frame to 2, under video settings on the notebook. This will now render every other frame.
Also reduce the amount of control nets you are using. Canny and open pose are good ones to keep.
it took me almost 5 hours for a 17 sec video😭😭😭😭😭😭
🫶
Could you please advise why someone should learn this as opposed to using apps like ’Loopsie’ or ‘Glam’ that promise to generate similar outcome?
Isn't warpfusion similar to using control net in stable diffusion?
Hey, it had the same control nets, but it gives you optical flow which gives a lucid feeling to the animation. This all can be done with batch in automatic1111 but you just would get the same quality/ look 👍
@@promptmuse thank you!
Can we do this with MacBook Pro M1 chip ?
im getting error with the text file saying that user_comment is not defined
Hi does stable warpfusion work on mac m2 chip?
Hi, does this work on Macbook M1?
Firstly pls tell what to put on the drive, then show the process, cuz when u run Colab, it eats hours. Thanks
unfortunately, google collab is not available in asian country, how sad, any suggestion aside from google collab?
An man, sorry to hear that. warp fusion is only currently available on colab. That may change as the development expands 🤞
Thanks it was really usuful. When I save my video and run the last cell it tooks almost 1 hour to complete though the video that I diffused(out put video) would be almost 1 second. I don't really know what is wrong.
Hey ! Jump on the Warp fusion Discord as its really helpful, the developer will point you in the right direction to making it run faster. There are new versions since this vis, so just check you are using the latest version of warp fusion :)
@@promptmuse Thank you:)
I got this Error if I try to make a video out of the frames in WarpFusion: IndexError: list index out of range. Any idea how to solve it?
Hey not seen that one but head over to their discord....its here : discord.gg/vXBTtzmeJd
I use the search function on the top right to search the bugs, there usually an answer, they are super nice over there as well. Good luck!
bleh.. youtube didnt tell me you were back making content.
haha had to take some time away as my body decided to fail again, but back at it again!
@@promptmuse You know, I can relate to this a lot.
I hope you are feeling better!! Your channel helped me start a mental health group! let me know if i can ever return the motivation.
hello sis,
i got this error message on 1.6 init main sd run function, cond_fn, color matching for SD.
ImportError: cannot import name '_compare_version' from 'torchmetrics.utilities.imports' (C:\code\WarpFusion\v0.15.4\env\lib\site-packages\torchmetrics\utilities\imports.py)
what to do ? thanks in advance
Sorry a bit late, but he’s still their discord link in my description. They will help you out with this error message in no time 🫶
Alrite, i'll check it later. Thanks so much.
Just to make sure, i dont have any knowledge about code, i hope the solution is newbie friendly.
Drag Gan is released, kindly show us a tutorial , thank you.
On it 🫡🦾
Google Collab to slow.. can we try it one runpod..?
Sadly it will be the same. It currently works on the google notebook with remote gpu and local. Warp is still in dev, but I’m sure will get quicker and quicker. Alex seems to be releasing a new build every few days 👍
Can i use low ram or low specs of laptop?
Sure can! It running from a remote GPU on google’s servers (That’s why it costs). You could do this from a phone if you wanted to ! 🫶
@@promptmuse thankyou, but i try the latest version of warp fusion , its getiing error,
can this be installed on pc with amd gpu?
You can switch to your own gpu on the colab notebook. It does depend how much ram you have though :)
@@promptmuse easier said than done, an amd gpu requires a bunch of workarounds that you can find online
i would like to confirm if it is can be used on a free colab or free kaggle account, if I pay you
Does this work on Mac M1 ?
Yeap !!
Darn is this possible on cellphone?
It is indeed, and an Ipad🔥
@@promptmuse thank u
copy path
nice tutorial. imo stable warpfusion should be open sourced its a script. lmao who charges money for a script? gen-2 is cleaner output wise if ppl really care enough to pay for a dev script rather than paying for server access
Can warp fusion run multiple GPUs?
You sure can, but not on the same account. 👍
@@promptmuse what do you mean?
@DJHOTCUE_ oh nice, what are you on pro or pro+?
Hey I am from India, I tried making payment of $10 but payment was declined. I am unable to get the warp fusion. Can anyone help 😢
Sorry to hear that. Google colab should be a available in India. You can download it locally which means it will use your computers power rather than google’s. For guide on this there is a discord link in my description and the guys who run it will help you out with any questions.
It would be nice to grab a spot of tea with you.
It's pronounced Tile V.A.E, variational auto-encoder
I made this music video clip using WarpFusion. Thanks to your help with this tutorial 🖤 ua-cam.com/users/shortsZy6YyCNZFvU?feature=share
can you help me to install it??? please
You are very beautiful, I was just looking at you.❤
Its free?
Hey, not free you have to pay for access (subscription the the developers patron, this is where you get the latest notebooks to download) . -£9
You also have to pay for google colab if you are not running this locally on you machine. -£10
Eu ensino a fazer pelo celular sem usar warpfusion e com o mesmo resultado, me sigam pra aprender
You have a 10,000 pound brain 😢😮
*I will seriously have a daughter with her*
Your tutorial is really, really great. I have followed the steps really close. But I still get an error when it runs 1.4. I have no idea what I'm doing wrong and why. I get the following message and tried to follow the instructions mentioned. I'm using Google Colab Stable Diffusion v0_15_7. And this is what it shows me when running 1.4:
ImportError Traceback (most recent call last)
in ()
10 import io
11 import math
---> 12 import timm
13 from IPython import display
14 import lpips
4 frames
/usr/local/lib/python3.10/dist-packages/torch/testing/_internal/logging_tensor.py in
8 from torch.utils.weak import WeakTensorKeyDictionary
9 import functools
---> 10 from torch._C._profiler import gather_traceback, symbolize_tracebacks
11
12
ImportError: cannot import name 'gather_traceback' from 'torch._C._profiler' (unknown location)
---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.
To view examples of installing some common dependencies, click the
"Open Examples" button below.
Any idea how to solve this or what I might be doing wrong?
Did you fixed the error?
Hi, i'm still getting an erro and can't get it to run passed Do The Run! Can you help? Here is my prompt and my error, can you please tell me what i'm doing wrong? a beautiful highly detailed man, ninja robot, warrior, dj, wizard with spirals, light eyes,
]
error: File "", line 2
]
^
SyntaxError: unmatched ']'
I think it's something at the end but I tried [}) nothing works...
can you pleeeeeeeeeeeease help me with it? i am really tired of it. i have a problem on rendering a video. if you give me insta, i can show you screenshot
U r awesome a r my source of income 🤫☺️🔥🔥🫵😉🤞🏼🔥🔥🔥