>How
Download the package and extract it on some fast drive that have good amount free space (100gb at least). Then
----Run 1stAIdBag_install.bat
--------Install ComfyUI and Manager. This is kinda bare minimum you need to install. You can install workflow's missing node via manager. Some user inputs needed during this.
--------Check if your ComfyUI runs after this. If it does....proceed
--------Install Custom nodes. This will install all custom nodes needed in my workflows, also other workflow's commonly used nodes (this might take some time)
--------Install Attentions. Highly recommend this as they give speed boost for generations.
--------Download essential WAN files and one of the WAN GGUF models (Q8, Q5 or Q2) Better your rig, bigger Q to get
------------NOTE: Downloading these models take time as those are several gigs
--------OPTIONAL: Download QWEN, GIT, or SDXL stuff if wanna play with image workflows

Run comfyui. On the first run it will take some time to load as some nodes like reactor will download some models needed. Once loaded, open the latest version of the workflow, re-select the all the models, vae, clip and loras. This is something you need to do every time you open up a new workflow, so the files point into locations in your computer. See the vid: https://screenrec.com/share/mkgZDHG5uI

After that, check other settings, type in prompt and hit run.

Buy me a beer: https://ko-fi.com/1staidbag

>Already got old install?
----Backup (cut) models and user folders somewhere
----del old ComfyUI_windows_portable folder. Then do as guided above
----paste model and user folders back into your fresh ComfyUI folder after install

>What do i need?
-Windows, RTX card, pref at least 8GB of VRAM, 32GB of RAM, drivers up2date
-You can run stuff with lower rig specs but then you would need lower models, lower resolution, lower clip duration, etc
-There are some info/help to run stuff with ATI cards in prev threads
-NOTE1: Major change on the pack, nothing is pre-installed, all nodes are getting installed via 1stAIdbag.bat now. Pack is now much smaller, 7megs

>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 (i'm not going to try fix other's workflows for you)
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)

Old thread(s)
https://archived.moe/r/search/subject/ComfyUI%20portable%20for%20newbies/

File: KP4_HB.mp4 (3.6 MB, 960x1280)

3.6 MB MP4

hello anons, trying to gen some high resolution SVI (using Wan Advanced I2V (Ultimate)), see vidrel

I'm having some issue where the video start acting weird after 3/4 segments, where the same sequences, prompts, loras, were doing fine for 6 segments

Is it because of the resolution (1280x960), where 1280x720 was doing fine ?

>>20531365
how about pic of that workflow....and i don't know what wan ultimate is (model, workflow, node?)...resolution isn't the issue, then that effect would be on all segments....first thing popped in my mind...did you accidentally load two high loras on that glitch segment?

>>20531437
Just checked, HIGH LOW in all the loaders. Renamed the loras with HIGH and LOW in all caps to avoid the mistake

Workflow here
gofile.io/d/C8ku1h

Wan Advanced I2V is from here, working better than IAMCCS node imo
github.com/wallen0322/ComfyUI-Wan22FMLF

thx for your help !

File: 2026-05-17 022159.png (216 KB, 777x530)

216 KB PNG

>>20531464
>Wan Advanced I2V is from here, working better than IAMCCS node imo
>github.com/wallen0322/ComfyUI-Wan22FMLF


who knows...since there is not a single line of english on that github page, what settings should do what. Setting wise it seems clone of iamccs. not going to install it....that workflow looks like hell....i mean my svi stuffed, but damn =)

anyways, this popped out to my eyes thou....i doubt that should cause the ghosting effect but who knows....and it seem that workflow has option to make video after each segment, use those so you can narrow down in which segment problem starts.....and if that workflow is from civitai, ask the maker

>>20531541
The thing is... it works well. I translated the page and it's the usual stuff.
Problem starts at the 4th segment, but the motion in the 2nd was already strange, like the lora was ignored.

I'll stick to what works, 1280x720 or 960x720, since problems started to appear when I went over 720p

Thanks for checking this out !

>>20531553
still say its not resolution issue as all segments are made with same dimension.....its something about the segment where have ulti-deephtroat loras loaded, do a run without them for test, if it looks ok, then those are the culprits

ghostdick guy here

thanks for your work, are there any wan 2.2 loras that can make or have the facial features consistent? i keep getting a whole different person

>>20531365
How did you generate the Input image? Is this inpainted?

How are people adding sound to their gens in Comfy? Seen a lot of good synched moans etc workflow?

File: KP6_HB.mp4 (3.86 MB, 720x960)

3.86 MB MP4

>>20532060
I saved it from the celeb thread. Don't even know who she is. vidral is the final version

>>20531573
Thanks for the help, I've regen the ending with a bit of variantion in the prompt and 2a couple of different seeds until I was satisfied. Turned out out good in the end.

File: 4ChanMahem_Audio_00075_.mp4 (1.81 MB, 704x384)

1.81 MB MP4

>>20532077

You have a few options for sound videos in Comfy: MMAudio, LTX, S2V, SVI2PRO and a few others I can't remember off hand. Each one has pros and cons and system requirements so you will need to do some research to see what fits your needs and system. LTX 2.3 was the easiest to install and best for what I wanted to do, attached is an example.

All my workflows are made from scratch but I am sure there are a bunch on Civi and most of the templates/downloads for these should be available on the Comfy manager.

>>20532086
couple days ago released dramabox is nice for fast voice cloning but it's mostly ltx.

File: PREcoverJan76new.jpg (349 KB, 1328x1800)

349 KB JPG

Is there someway to rearrange the 6 svi nodes, because trying to skip one doesn't work the steps must be numerical order. It is painful to start from scratch, changing loras and prompts for each segment. I've been trying to figure out what the triple sampler does since the 1st svi wf. Does it increase speed or quality I'm not sure.

File: images (10).jpg (13 KB, 240x240)

13 KB JPG

Got a nice glow going making a picture from behind in sdxl. Then on svi segments
1 segment follows the woman from a hallway to a bedroom where she gets onto the bed, camera follows focusing on ass, other segments are self explanatory and ends woth bodyshot lora (pain to get working wording)

Flow* not glow. One of the best prompts is following the woman up stairs view from below ass focus.

>>20531852
not really....this been asked many times. Models and loras always alter the face and body, they are not build with your specific input woman in mind so they have work the dataset they have and that one single image you use as input. Some loras have better datasets where the author have spend some time to blur the faces in images/videos he is using for lora training --> generally these have better face consistency. there are plenty of loras for wan that do the same thing, some deepthroat lora works better than other..

stuff you could is try is to lower the lora strengths, especially low ones (those generally have texture data in them). Other way to "force" consistency is to use first frame / last frame.

>>20532077
like other anon said....ltx is the easiest way to include audio in your gen. mmaudio can add audio to your existing video where it looks at your video and tries to makes sounds based on that (plus what you prompt)....s2v is where you have audio file first, it makes a video driven by the audio (plus prompt). it needs its own wan model

>>20532274
if you mean skipping some segments between....no, as it needs the latent noise from previous part. AND when i say no, i mean it MIGHT be possible with somekind of wizardy and time.

>It is painful to start from scratch, changing loras and prompts for each segment
Well.....duh? That's the way shit works. I know it is frustrating, but it is also rewarding you you've finetuned your loras and prompts for all of your 6segments.

>what triple sampler does
it adds few steps into the beginning of the sampling without lightx loras and higher cfg value. This is to get better prompt coherency and better motion (since lightx lora limits those aspects alot).....its much slower method, for very minimal quality gain. I've removed that from my wf and moved back to simple 2pass method

File: fd5eb11979fd72eae02cc44a1(...).jpg (98 KB, 736x1104)

98 KB JPG

Critical useful info declassified

File: marla monroe taking you t(...).mp4 (1 MB, 352x640)

1 MB MP4

Sdxl and wan wf
Starting Image https://files.catbox.moe/t1auew.png
Svi video
https://files.catbox.moe/ygvc72.mp4

File: 2026-05-18 014236.png (718 KB, 2484x911)

718 KB PNG

>>20532621
there you go =)

but....i dunno why you have so many loras loaded that has no need in the segment? you don't need doggystyle lora loaded when you prompt "woman walks up stairs" or cumshot lora when you prompt her to kneel down

even if you don't use the loras (like prompting the keywords)....they still get loaded in with the model, bleed their stuff into the noise pool and therefor affect your outputs.

keep it simple...don't overload stuff with loras.

File: 014523_157356776554552584(...).mp4 (2.62 MB, 432x768)

2.62 MB MP4

>>20532679

>>20532688

Shoulda had a jumpscare going into the room. Thats a total horror cam shot lol

File: hqdefault (1).jpg (19 KB, 480x360)

19 KB JPG

Thank you Dr AId and yeah it is but it's originally inspired by Stellar Blade climbing a ladder. I had the loras at 0.00 and in the console they're not getting mentioned only when above 0.01. strength. Makes sense about a noise pool, to be honest senpai.