img2vid Motionmuse.ai

T

Tommy9x

Lurker
Joined
Mar 24, 2022
Posts
15
Reaction score
0
Has anyone reversed engineered these templates? I've used it briefly and it produces amazing results. I currently have a fully functional wan2.2 img2video workflow working, but I was wondering if anyone had an idea on how to create these type of videos found in motionmuse.ai.
 
Yes it very easy to make on muse but it only give 5 token Every day and the minimum requirement for all the effects are 30 and you can even use multiple emails it detects even if you manage to make multiple account and claim the daily rewards it will not allow to get the daily rewards in other account 1 per account daily we need a solution for that
 
Sorry man, I'm not asking how to use the website. I'm asking if there's a workflow that I can generate myself since I have the hardware for it. What models they use, loras, etc.
 
https://motionmuse.ai/train-tutorial i just found this so you just upload videos, photos with a prompt. Then they train the template for 30 hours. I think people upload the same video multiple times and thats why it looks the same every time and it just snaps the woman into position. I feel like it should be possible with Wan Vace but i dont use it so idk :peepoShrug:
 
I’ve tried wan vace. It just takes your starting image and models it to the movement of the video. I’m thinking it’s a mix of models/methods. Cant quite figure it out.
 
Is it possible to get results like MotionMuse locally with 8gb VRAM? I'm wondering if I went through the hassle of setting something up like Wan 2 or whatever works with 8gb, would I get what I see on motionmuse? I'd be fine with it taking longer up to like 5-10 mins so long as the results would be good like on motion.
If so what should I install locally to make that happen? I already use comfyui here and there but haven't done any img to video so far.
 
For 8gb vram, use gguf models but honestly 8gb is really minimal for img2video. Need at least 3080, 4080. Your generation times depending on resolution and frames, generation can be anywhere from 15 minutes to 55 minutes. Visit CivitAI for guides. You won't get exactly like motionmuse, it will likely look a little blurry and results will not be as consistent. Motionmuse is really clean which is why I'm asking if anyone has a workflow that can generate that type of video. I am close with a wan 2.2 workflow but I'd like it to get cleaner.
 
im currently working on a motionmuse script that gives 9999 credits so far it almost works says generating video but doesn't actually do it so if anyone is good with scripts and wants to help that would be great
 
is it actually possible to get results similar to motionmuse for free if you've got a very good PC ? i've started looking into comfyUI and wan but its all quite overwhelming
 
Motionmuse uses wan under the hood, so its literally the same thing, i just use wan on tensor (with custom loras) or higgsfield
 
i guess what im asking is can you use wan locally for free or do you have to pay for it too ?
 
Likely won't work. What you're seeing is client side number edits. Motionmuse credits are almost certainly server side. You won't be able to edit those values and highly likely a waste of time.
 
For motionmuse, it looks like it takes the first second of the image and generates a flux to use as the base of the next 4 seconds. It's that transition that I'm trying to figure out.
 
I am curious what loras they use for the "cum mouthful CUT" template. That one is excellent
 
the telegram option is so expensive its crazy but i guess it’s worth getting it, the ai generation is really nice
 
It really is convenient. Here I'm hoping for someone to come up with a script that will make all this coin-farming less painful or something along the lines because it's truly expensive.
 
use crypto its much cheaper, i didn't know anything about crypto too but i figured out in like 10mins
 
You must log in or register to reply here.
Back
Top