Thread 20053516

:::Apprentice pack (v0.3.60):
::::::::https://mega.nz/file/3h4y2KgC#ZQsoH7JHyzb5StA64kiUTVByMkhkc5Byro24DP5M548
::::::::https://mega.nz/file/j5pD2bwQ#cbqilnmiePksu4XSIM6t91EJ7XlLemQ4yAYb6ldrPZE (1staidbag.bat fix)

Updated workflows since initial pack release:

::::WAN 2.2 v1.12 workflow (blockswap added):
::::::::https://mega.nz/file/fsZVAapK#T1g8SUlgELHpSitWr6iVqfLXTZpjIadmsX32ExL8aHo

::::WAN 2.2 v1.13 workflow (MOE/Auto-split sigmas sampler):
::::::::https://mega.nz/file/zoQWGJ5B#zcQojnO1IDUOB2xWSvn0fdf_MdAtJce91yaohMbhX2w

::::WAN 2.2 v1.14 workflow (StartFrame/Endframe option, NOT VACE):
::::::::https://mega.nz/file/GsgD0Krb#400s829w2xrjK5KpjI0w4mYTFDscc57nAvLJf_nDwKE

::::WAN 2.1 v1.13 workflow (based on 2.2 one but single sampler):
::::::::https://mega.nz/file/ilhmARTL#ZvV93hRQA7XojDhpcTBSkC70BClHI2ARd8auQTe3GNE


Old thread(s)
vol1 https://archived.moe/r/thread/19910536/
vol2 https://archived.moe/r/thread/19938024/
vol3 https://archived.moe/r/thread/19987607/
vol4 https://archived.moe/r/thread/20011173/
vol5 https://archived.moe/r/thread/20035627/

>What is it?
Modified version of comfy portable
Aimed for new apprentices who wanna jump in AI world. In the package
----Kijai's WAN-Wrapper nodes pre-installed
----Pre-installed nodes needed for my wan2.2/wan2.1 workflows (wf included)
----Pre-installed nodes needed for pisswiz's wan2.2 workflow (wf included)
----ReActor preinstalled, NSFW patched (note that updating the node might fuck it up)
----Sage & Triton installation
----Simple batch file by me that downloads all needed files (vae/clip/lightx-lora/ggufs) to get you started
----Incl. program called LosslessCut to join your clips together
----Couple upscale models included

>How
----Download the package
----Extract it on some fast drive that have good amount free space (100gb at least)
----Run the run_nvidia_gpu and see if you get your comfy running
----....if so, good, close it up. It won't do anything as you don't have models yet
----Install VC_redist.x64.exe and reboot (most likely you have this installed already, needed for triton)
----Download (1staidbag.bat dload fix into your comfyui root, replacing the old one, fixes lighx lora dowloads)
----Run 1stAIdBag.bat
----....Download options 1&2 and pick one of the GGUF models (Q8, Q5 or Q2)
----....Downloading these models take time as those are several gigs
----....Install sage and triton (also updates flash attention)

>Already got old install?
----Backup (cut) models and user folders somewhere (pref same drive's root where your comfy is)
----del ComfyUI_windows_portable folder
----read how section above
----paste folders back into your fresh comfyui portable

Once all downloaded. Launch Comfyui using run_nvidia_gpu_sageattention.bat, RESELECT ALL MODELS, to make sure
they are all there in your computer. Load image, type something in the prompt, run.

If expecting path/string(NoneType) errors ---> you didn't reselect the models
If sage related error = you didn't install it correctly/ or at all --> install sage, or disable the nodes
If out of memory / allocation to device --> lower output resolution, shorten vid length, lower steps, enable blockswap

If/when you need help....
A) take a screenshot of your whole workflow (and log console window if possible)
B) state what workflow are you using
C) what specs on your rig

>Where to get loras?
https://civitai.com/search/models?baseModel=Wan%20Video&baseModel=Wan%20Video%202.2%20I2V-A14B&baseModel=Wan%20Video%202.2%20T2V-A14B&baseModel=Wan%20Video%2014B%20t2v&modelType=LORA&sortBy=models_v9%3AcreatedAt%3Adesc
(you need to make account and enable show adult content)
https://civarchive.com/search?type=LORA&base_model=Wan+Video+14B+i2v+480p&is_nsfw=true&sort=newest&page=2
(old 2.1 loras that got removed from civitai due to their nsfw bullshit policy)

Is there a specific lora or prompt you use to undress quickly?

Any non-mega links? I'm already at their download limit and don't want to use a proxy

>>20053712
i've also struggled with this. prompting to remove clothes might take up 80% of the video.
'clothes disappear' or 'she removes all clothes' or 'with one swift movement she pulls all of her clothes off'
sometimes it just refuses to work no matter what

any nsfw lora seems to help

>>20053712
>>20053979
try "scene suddenly change where girl is naked"

>>20053759
yuo can resume it after few hours

>>20053999
yeah if i prompt that she is in a completely different environment from the picture, there is sometimes a sudden change - sometimes it's a strange picture wipe effect, other times there is a fast fade transition. yet to work out how it decides which it goes for.

>>20053979
if you only care about getting them naked, i like to throw this in front of my prompt to just get the clothes off,

the woman has a clothing malfunction causing the woman's clothes fall apart at the seams. clothes falls off her body and onto the ground quickly, exposing her naked breasts and naked vagina and completely naked body the woman is completely naked leaving her wearing only white sneakers. then

it basically makes the clothes fly off the body, so it doesn't waste time trying to take it off then to do what you want

yes it does have a lot of the same prompting on there, it's for confirming prompts if that makes sense

>>20054098
naked with white sneakers, a man of fine taste

What is the difference between Lightx2v and Lightning?

>>20054165
i'd say nothing if you are talking about the loras...same lora, different names

>>20053516 (OP)
Have you been using wan animate? Wouldn't it be better than the RNG we generate normally?

>>20053522
>enable blockswap
finally it works

>>20053516 (OP)
Do you do a T2V pack ? your workflow is one of the only ones I've managed to get working.

>>20053522
just a question, what's the best sage option for potato pcs?

getting these hoes to strip is so cumbersome,any decent add on for it?sometimes>>20053979
this might help but weirdly it also zooms a ton on the bellies(?

for boob reveals in the wan2.2 1.14 workflow, i'd use I2v-tittydrop-e26 for both high noise and low noise? is there one specific for wan2.2?

Just a quick question.
I'm currently generating pictures using

wan2.1-t2v-14b-Q8_0

Is there an improved NSFW specific unet GGUF to create porn pics?

>>20053522
Which lora is this and the workflow is quite successful can you help me

>>20055405
If you mean something similar to some sdxl checkpoints.....no, or at least haven't seen one. So just the base model and loras is the way to go

There is specific nsfw Wan text encoder that should do better with xxx prompts

up

>>20056370
Thanks!

Where do we find larger than Q8 GGUF models?

>>20057347
Q8 are the top gguf models, if you want better quality then use the diffusion ones fp8 (quality almost the same as q8s) or better fp16. Note that you need better rig to run these, they also eat more time to gen

>>20057361
https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models

>>20057361
>>20057374
Gotcha!

Is there any way to increase video quality by sacrificing the time it takes to make it? I don't care if it take up to an hour as long as the quality is good. Or will it just run out of memory and fail?

>>20057516

what's the point

>>20057731

Because i dont have a very good card

>>20053516 (OP)
How much are the pc requeriments to run this?

>>20057516
use better models, increase the reso, increase the steps, don't use acceleration loras....but if you don't have good card/ram amount you will face OOM pretty quick

try increasing the output resolution (need to have good resolution input image of course) and steps little by little at first

>>20057959
i'd say min = rtx 30xx and 32gb....lower can run it but very slow and less quality (as you would need to use lower models)

>>20058000
question not same persom, how does sage help and what are the best ones to use for lower end shit. For context have a 20x, generates at the low end settings and generates just fine, just find it slow. heard sage can increase speed

>>20053519
How much VRAM do I need in to use Q5?

up

>>20058245
i've always left it to auto...there is not that many options if you wanna test them out. sage 2.2 should boost the speed about 20-30%

>>20053516 (OP)
just wanted to say thanks, this helped accelerate my understanding of how this all works. still a long way to go, but a great springboard.
got any loras you like?

>>20053516 (OP)
Anyway to get your StartFrame/Endframe workflow to output perfect loop vids? the loops it give are a bit choppy.

Up

>>20059158
never planned on using that for loops.....try adjusting/disabling the wan skip last frame node...... i dunno really, would make sense to skip the last frame if it is the same as first frame thou (to avoid stutter at the end

I tried this and keep getting this error:

Power Lora Loader (rgthree)
Error while deserializing header: header too small

# ComfyUI Error Report
## Error Details
- **Node ID:** 103
- **Node Type:** Power Lora Loader (rgthree)
- **Exception Type:** safetensors_rust.SafetensorError
- **Exception Message:** Error while deserializing header: header too small

I didn't change the workspace at all, everything's the same as yours when you uploaded it.

I cant figure it out, pls help

>>20060146
download the lightx loras manually into models/lora folder (my auto-dowload is broken in the pack)

high: https://civitai.com/api/download/models/2361379?type=Model&format=SafeTensor

low: https://civitai.com/api/download/models/2337903?type=Model&format=SafeTensor

>>20060158
oh and once dowloaded...restart comfy and re-select them on the dropdown list (they are newer versions)

>>20060160

hey thanks for the quick response, it seems to be working now from my test run!

so, to get add new loras, I click "Add Lora" in the "High noise LORAs" node? I still need the lightx loras, right? I just put the new loras under that one?

Also, this works for pictures of all sizes (resolutions) right?

Also, any way to change the length of the video?

Sorry I'm brand new to this

>>20060172
no need to add new lora.....click the existing one and select the one you just downloaded (on both sides, high and low)

>Also, this works for pictures of all sizes (resolutions) right?

yes...input image sizes won't matter, they got resized before the gen.....some say 1:1 pictures work better, like 720x720...but i dunno about that on wan 2.2

>Also, any way to change the length of the video?

there is a slider for that in the workflow...keep it in 4-6 secs thou, wan can't handle more (atm)