Thread 19782452

From what I gathered, looks like everyone is using ComfyUI with loras and all the stuff to make the magic happen. But other alternatives are sites like tensor.art with some templates.

What is even out there that people don't know much? Feel freee to share the sauce to other methods.

I believe under Tensor art templates, there are actually comfyui template running under the hood, those templates just come preloaded with everything
needed like models, loras, working settings.

For videos, comfy is only reasonable option. For images there are some ok app like fooocus, automatic1111 where learning curve is so deep as in comfy

>>19782477
Thanks for the reply, bag.
Yeah, I've seen that you can customize a little bit in Tensor. But nothing comes close having the latest GPU and dabbling with comfy/loras.

How it was the learning curve for you in the comfy method? It took one week to get the hang of it?

>>19782452 (OP)
what model on tensor.art should i use?

>>19782483
This is the templates I've mentioned:

This video has a man stripe a woman
https://tensor.art/template/run/869638843001747904?source_id=kTm1oVzgkEC0rPEpaH388BAl

This one pulls up clothes:
https://tensor.art/template/run/869462680623131189?source_id=kTm1oVzgkEC0rPEpaH388BAl

cum shots:
https://tensor.art/template/run/869453790040879609?source_id=kTm1oVzgkEC0rPEpaH388BAl

blowjob:
https://tensor.art/template/run/869586719278658195?source_id=kTm1oVzgkEC0rPEpaH388BAl

>>19782482
Still learning.....=) i mean you can get the hang of the basics pretty fast, what node connects to what, ie, if you want to upscale, use this node between these etc...where to put models and model patchers. Only downside on comfy is sometimes you need to hunt down some nodes that are not avail on comfy's node library. Other thing is that when you update it, most likely something breaks on your workflow. But you'll get use to it =)

I used to to download multiple workflows and look how they work, then starting building my own based of the best bits. There is lot going on WAN related side, acceleration loras, new guidance nodes that update almost weeks. You can get pretty awesome results just in few steps even on lower end cards nowdays if you build your workflow right....example vid you posted on OP took me 10 steps, about 5mins on my rtx4070 (well 3 x 5mins as it has 3 clips joined together)

>>19782619
The part about hunting nodes must be a nightmare. Because some people complain about having a great GPU and not having good results. I guess it's all about the right connections and libraries.

I've seen a lot of mentions of WAN on Tensor. Is it a person that makes lora/libraries/workflows? And how easy is to create a preset of the entire workflow to share with someone? I'm trying to find a guide to beginners but this topic is pretty scarce for guides. I guess I need to search on the deepweb or something, lol.

>>19782633
Well having decent GPU is the most important thing ofc you want to run stuff on your own computer, there're are alternatives ways to run comfy like cloud services, they are way faster, but costs some money....

WAN is the name of the main video model. Other popular one is Hunyuan, but most people use WAN nowday, quality is just that much better.

Loras are kinda like extension-packs for that main model that are trained to certain specific thing/action very well using sample images or videos

Workflows are easy to save and share to anyone. Other guy just need to have the same nodes and models installed and they are ready to go. Comfyui is pretty good nowday on getting them for you. Example, i'll sent you my workflow file (,json) and you open it up in your comfyui, it checks what you are missing and there is basically one-button-press to install them all.

But reserve some discspace....just by all main-and-necessary models and some basic loras easily eat up 60-80gb from your hd space, just to get it running. There are loads of comfy tutorials on the net/jewtube......no need to dive into deep for them. Watching tutorial on how to make cat dance on WAN isn't any different than how to make girl pull her top off....process is the same

I'm trying to use it with cpu. But I'm not finding the models for titdrop, nudifying or stripping. If anyone can help me with that, I'd be thankful.

>>19782807
civitaiarchive com

>>19782809
>civitaiarchive com
yeah, I'm searching in it. But do you have any indication for low RAM cpu? I know that my creations will be very limited, but that's all I can do for now. I think I can't use other base model than SD 1.5.

I'm a retard when it comes to all this, so is there one giant thing I can download that has everything I need? Pinokio kind of did that but it's having issues now

>>19782935
I don't think you will be able to run on only CPU. It can offload certain things to CPU to save VRAM, but it needs a GPU.

>>19782935
well i have 32gb of ram.....that is slow, but it gets the work done with tine....i assume you mean VRAM (your graphic card ram)

lets say you have RTX card with less than 8gb....dont bother. I mean you could run with now that there is new loras, but you would have to use some smaller Q(auntisized) models, but quality will look like shit. Native WAN model takes up 32gb of your vram to run straight up and fast, but no1 has those cards, so therefore there are nodes to help in comfyui. On 12gb card you can make shit like i do (and yes yes, i will release my workflow soon)

>>19782948
i first discovered WAN via pinokio as well, felt left from control, so i moved to comfy

>>19782487
do you know where or how could i find actual sex models in tensor instead of just bj models?

or any full nudify models?

why does verification keep failing on tensor?

do i need to enter id/bank/payment info for it to work?

>>19782487
How do I find more like this? I already hit my limit on all of these and don't want to subscribe. The in site generator keeps throwing errors, but these drag and drop picture ones work just fine.

>>19783105
>>19783120
>>19783140
These are the reasons why all use comfy ;)

WAN can handle nudity all ok.....how to get people nude --> needs loras

>>19783190
how do i get loras tho

>>19783201
civitai or civitaiarchive (for deletes ones), or huggingface or tensor art......loads of places

>>19782452 (OP)
Akool and pollo.ai for most stuff celebs etc

I'm using motionmuse

motionmuse.ai/r/w6aosada

>>19783289
are there nudify models on pollo.ai?

>>19782487
>>19783187
I’m looking for more links, I’m in the exact same boat in the sense that I ran out of free tries – limit was three – but I had no luck searching for the same elsewhere.

My ask on the discord also fell on deaf ears, we need to build a list / archive of tensor links.

What do I used if I want to make photos or videos of my ex getting a BBC gangbang. Don’t mind paying

Does anybody remember wanker.ai?

the UI made the entire process really nice / simple.

You could just upload an image of a face, and choose a set of parameters (facials, handjob, body type etc…), and it would generate a new image of the girl in an apartment / context and the results were pretty great.

Is there a way to do that now that the site is gone?

>>19783206
Hi 1st.AId.bag, any chance you can send me your comfy video .json file? I'm always blown away from how real your generations are

My discord: .tdawgz

>>19783190
So it's impossible to make 'fake' tensor account with temporary mail ? How do other makes multiple accounts at once ?

>>19785090
well no, i think people do that

>>19783767
i will shared once i get it done, need to change some nodes and clean it up

>>19783627
Yeah are there free web nudifiers still around? These were all good but they're now dead

wanker.app
nudigen.com
deep-nude.ai

>>19785300
I found wanker.app to have really good results. The likeness the site achieved was excellent. I’m not sure how they setup uploading single or double image to achieve better results than some loras I’ve seen on here.

Would love to figure out how to do that in Tensor.

>>19785295
Sounds great, looking forward to it!

>>19785392
whats the url for wanker.app?
i just tried wanker.app but it didnt come up

>>19785908
it's already dead

looking for a free online replacement to nudify and gen

preferably no signup. a few daily limit is fine

>>19783545
I believe you can

Bump for gooners to the front page. Try this instead of begging

any other sites?

>>19789002
Nothing can come close to creating them locally

>>19782948
I suppose you want a video venerator all in one.
An image generator tutorial all in one would be hard to slap together.

4-step run (rtx4070 12bg vram, 32gb ram)

I'm not getting free credits on akool anymore. Was that killed?

>>19789960
spoonfeede json plz

>>19789988
what kinda setup you have....would be nice to test on lower rtx30XX cards

>>19789989
I'm not the same guy but I got a 5060ti and I'd love to be a json guinea pig

>>19789989
3060, was thinkng about redownlpading a portable version and restarting from scratch.

Do one bald, then try again with sage attention and triton, as the internet implies their installationis finnicky

>>19789995
got disc? would like to see if can get if working off the bat

>>19790004
i did sage/triton on clean comfy portable on this guide; https://www.patreon.com/posts/124253103

is 3060 8gb? also if you have disc, lemme know, i've added blockswap into my workflow, never needed it so wanna know if it has affect

>>19790008
12gb, and I technically do, but I only ever used it once for a gacha game that got me addicted but finally died and a a few other minor game related thingers.

Any way to share that isn't public? I can give it a go later tonight.

>>19790006
Yessir, my username is: .tdawgz

>>19790013
alternate would be this temp email Injust setup.

If that's not good enough, that's cool, I am proabbly being overly paranoid, and the alphabet people could probably find me through these posts anyways, but discord just seems more personal than an anonymous imageboard.

hexayi2834@axcradio.com

>>19790014
>>19790016

im just generally interested if after load ppl can get nodes off from comfy manager...kinda my first own workflow and, long doing and get remember if some node are via git

>>19790013
>>19790022
TriangleCircleSquareCross

>>19782487
how do you get thoses, i already suscribed to the profile but i cant see the other tools

Still looking for some comfy-faggot with lower tier card (8gb) to test the workflow out

>>19790837
i have comfy ui already installed with 6gb, 1660 card, but i didn't test it yet, im learning

>>19790984
with a 1660 you won't come far

are there any good online other then tensor? its easy enough to make lots of gmails but it gets old logging out and back in

>>19790984
i think with low Q model you could get it working, but it would be loong wait for shitty quality

Anyone with a good guide on how to make Loras from Civitai(archive) work on ComfyUI?

I'm trying my hardest to get image to video working with loras, but I cant wrap my dumb head around some things.

Load diffusion model, load clip, load VAE, Load CLIP Vision: Why is it undefined? Why am I so retarded I cant find online guides on how to make the undefined go away?

>>19793835
post a pic of your workflow, are you sure your loras are in models/loras folder?

>>19793841

On the left side: What are all the nulls and how do I get rid of 'm?

And how do I get the Lora to work in general?

>>19793841
Is there a Chance you could Upload the entire Workflow you Posted? I Can’t see every step - your work is absolutely brilliant!

>>19793851
what is BBC.savetensor? note that you can only use WAN loras in WAN workflows. Those null(red) means the node didn't find any needed models/files from the folders they should be.....

main model into diffusion_models folder
vae file vae folder
clip file in text_encoders folder (set type to wan in your node)
clip vision file in clip_vision folder
loras go into loras folder

once all in correct places, restart comfy

when you save loras into your computer, add prefix to them, like "wan_loraname.safetensors"

>>19793887
Dear 1st Aid Kit,

This is Gesetzgeber. Over the past few days, I’ve perfected my facial workflow and have already helped many users with it. I’ve seen some amazing and brilliant work from you and would love to exchange ideas with you—maybe we can help each other and refine our workflows together. Is there a way to get in touch with you? Looking Forward so we could help each other.

>>19793903
post your disc and i can get in touch

My Disc is gesetzgeber

>>19793913
other source for loras for video AI that isn't civitai?

other source for loras for video AI that isn't civitai?

>>19794022
there are some in tensor art that arent avail on civitai...i dunno where ppl will start posting their nsfw loras (that are againsts civitai's new policies)

bumping for more free online tools. tensor is great

Anything that works with AMD?

>>19794183
Properly will re-register and not pay for it everytime lol

>>19794221
Image gen or vid gen?
I had Comfy working for image gen on a RX580 before an upgrade.

>>19794221
google zluda....ppl have got shit running with amd cards, not sure about the speeds or quality

>>19794342
Hello almighty wizard, do you know the best way to face swap someone in a video?

>>19794370
Facefusion is what use

if you use motionmuse, sign in and dont use credits, then check in the next day. dont use multiple accounts, theyll actually penalize you and restart the timer on you.

Has anybody ever made a Lora of somebody they know in Tensor? Was it difficult? Any tips?

>>19796154
you need to run lora training on pretty high end cards like L40s, so cloud running is pretty much the only option

you need decent amount of images (dozen at least) plus you have to desc/prompt the every image. This is most work you need to do, but better descriptions you and higher quality images you provide pays at the end

Once dataset is prepared, you load it up on diffusion pipe.....wait (and pay for) 4-8hours (depening on settings you use), and keep your thumbs up.

But yeah, i wouldn't say this is the first thing you wanna try out when entering ai-world

>>19796249
Legend, thanks for your reply, truly.

I’m using tensor about a week or two (with no prior experience), and slowly figuring it out, and have been making gens with some success, but mostly failures. The Remix function on their site has been where I learned the most about Tensor.

I tried to make a Lora of somebody I know (basic youtube tutorial) and they are coming out pretty okay, but when I try to add additional Lora models (for skin etc…) the gens look like shit, irrespective of the prompt. I can make okay gens with other peoples Loras. I find making good Loras it is the only thing holding me back.

As a complete newb, where can I load it up on diffusion pipe?

>>19796249
M8, going through this thread you have restored my faith in other wizards. seems like a lot of them are keeping this stuff to themselves. you are a legend for helping others learn.

I had to figure most of what I know on my own and yet still have trouble with the basics of I2V. even that image you posted detailing your loras is a massive help to me. I can find specific loras when searching tensor site but I cannot for the life of me filter search for all the nsfw ones in a single page.

do you use a specific shift/cfg/steps when doing different things /facefuck/tittydrop or do you use one setting for all. would love to know as im currently flicking them up and down trying to find the right balance.

thank you very much for your service here.

>>19796249
for video?
for sd 1.5 i remember using dreambooth on google colab and it only took a few hours.

>>19796860
i think you can cook up first version epoch in few hours...most likely not that good. Never made image loras, but i would assume they would be faster. It kinda comes down to how big your dataset is

>>19796813
if you use single sampler setup....shift:5, cfg between 3-5, steps 15-30 (i found that 20 is pretty decent enough quality).

if you use dual sample setup with accerelation loras, cfg must be 1, shift 8-10, steps can be low as 4-6. Downside is; it will loose lot of the movement

>>19796705
runpod or mimicpc are good options

Sure this has been asked already a thousand time but assuming none of this is possible with a macbook

>>19798166
First results in google imply it is possible, though I have never done it.

>>19798166
like with amd cards, with some tweaking and knowledge, i guess it is possible. Getting comfy running on mac seems no problem, but how to get your graphics cards to work with it might be biggest barrier

8gb RTX3070 user looking to get it working. Learning comfy after a couple years on automatic1111.

My disc is throwaway421

>>19798069
Thank you very much!

>>19785300
This one is like the other ones and haven’t been nuked yet
motionmuse.ai/r/b94tb8o7

>>19783206
Major update, Tensor WAN Loras got nuked.

Making two servers for genners and gathers soon.

>>19800823
oh wow, really?

>>19799836
This is a sneakybeg referral link fyi

Motionmuse has okay gens, not as good as tensor. Theyre a pita to work with though and you only get 20 creds to to start. Daily login is 5 creds and it costs 30 for anything nsfw. they also track logins and will deny daily creds if you log in on diff accounts too

>>19782487
Do you have a deepthroat or face fuck one?

>>19801265
paying money to use foss

morons, all of em

>>19782487
can you or someone else check and see if that creater named 『AI』£1N has anything else or is it just those 4

>>19782452 (OP)
Does anyone know any free software that's easy to us to upscale a video? All i need to do is upscale from 406 to 600 and something. I haven\'t been able to find any web browser based ones that were free so i figure that my only option is software.

>>19800823
yeah, very sad

missed downloading a few loras, hoping someone has backups

>>19801383
theres this, but its not great https://openmodeldb.info/
how long are the videos?
post a sample clip if it doesnt violate your privacy?

>>19801633
This is one that I'm trying to do. I think it's better if I learn to do it myself (hopefully it's not too difficult) then try and bother you guys here. Thanks for responding to my request, man.

>>19782487
i did one pic and now im getting an error from all 4 links? why is it doing that?

>>19802953
All nsfw stuff got shut down. pain and sorrow.
Anyone got an alternative?

>>19803009
they cant control the net tho, wont be long till some russian site opens up and gives access to all nsfw loras. they cant stop it lol

>>19803093
Tensor is one hosting site. NSFW loras have been available, and are still available, on literally dozens of other sites.

lrn2google summerfags

>>19803102
no one is saying they are banned everywhere m8, dont stress out lol

>>19803107
https://pollo.ai/invitation-landing?invite_code=iWQ4J1

>>19803170
i get free credits if u click on link

>>19803171
motionmuse.ai/r/zxva0x8

>>19802861
i used REALESRGAN_X2, you can download it from https://openmodeldb.info/
i used comfyui but you can use chainner if you want.
i also doubled the framerate if you are into that sort of thing

>>19803425
thanks man, that was kind of you. it was a bit overkill though as i feel that upscaling too much starts to really show how fake it is. anyway, if i may ask you a question? do i need a powerful laptop to do what you did if my only goal is short clips like these? i doubt that ill ever go over 15 seconds. my computer is pretty shit and about a decade+ or - a couple years old.

im just too stupid for these things is there no online interface for img to vid?

>>19803501
depends on how patient you are.
ive done this in the past with a 6gb nvidia gpu, might be possible with lower vram, dunno

>>19800823
>>19803102
Noooooooo. Tensor was the only way I was having fun without messing with comfy/GPU. Do you think it's possible for a comeback? Or another alternative? Fuck, Tensor was so good. Why the hell they had to nuke it.

>>19803551
Here's their statement:
https://tensor.art/event/NSFW&CelebrityAdjustments

>>19803541
yeah, my system is much worse so i guess im out of luck lol. thanks again though.

Anyone got a guide or tutorial on how to setup and use wan 2.1 in comfyui to generate img to vid from scratch

>>19804056
1. Have 8GB VRAM minimum, hopefully Nvidia (because it is better)

2. Download ComfyUI portable, unzip.
https://docs.comfy.org/installation/comfyui_portable_windows

3. Install ComfyUI manager
https://github.com/Comfy-Org/ComfyUI-Manager

3.5 Install Triton and SageAttention
https://www.patreon.com/posts/124253103

4. 1st Aid Bag's other thread and download his workflow and install the nodes via ComfyUI manager. His workflow has links and notes for all the other bits and bobs you need.
>>19799683

5. Then fiddle and read and lurk and beg for more information until it works.

I probably skipped a step or wasn't very clear since this was off the top of my head, but you won't get far unless you have some troubleshooting and critical reading and thinking skills anyways.

bumpp

pretty solid for free gens

motionmuse.ai/r/dtaopigq

>>19804666
how often do u get free tokens?

>>19801138
Yeah, figured that was sensible enough.

Since Celeb Loras are banned, and I'm still hunting them down.

>>19803425
If you have more clips of kaylaagabrielle, I'm all in.

disc:channybae.