Overview
A female bodies expanding into serious thickness. This model focuses mainly on thigh/hip growth, breast expansion and ass expansion, but can also do other things. Just play around with it.
There is no trigger; use following descriptions:
Her thighs grow (massively)
Her breasts grow (massively)
Her buttocks grow (massively)
Her belly grows
Her height increases (dramatically) / her head hits the ceiling
Her clothes rip / tear / shred / tear open / rips across / leaving exposed
Adding my T2V helper LoRAs will improve the quality and/or growth rate significantly. Also, using the TripleKSampler can greatly improve the animation quality, if you're using speedup LoRAs.
Further descriptors can enhance the effect. Sometimes it can take a few tries to get it just right. These keywords can also be used in the negative prompt to weaken an effect.
For more have a look at the example images/videos.
LTXV 2.3 - V1
Coming in late May/early June...
WAN 2.2 I2V - V2.9
Training model: WAN 2.2 I2V
LoRA strength: 0.5-1.0
Sampler: euler / euler a
Use TripleKSampler with 1/4/4 steps or no lightning LoRAs at all for quality.
I highly recommend using my T2V helper LoRAs linked below. Use their low-noise component for improved texture quality or their high-noise for better shapes and bigger expansions.
A woman in a bra standing in a bedroom. Her breasts grow massively; her breasts become gigantic. Her bra stretches tight. She looks down shocked.Description
FAQ
Comments (38)
Quick question ; did you forget High and Low in the filenames or did Civit remove them and replaced by those 4 letters? Cheers
Thanks for the comment! CivitAI somehow completely changed the file names, sorry. Maybe I can still fix that.
@cupra Yeah Idk why they did this, all filenames now end with 4 random letters somehow
When I go to "Manage files" the safetensors are still called "wan22_i2v_expansion_v2.5_high_noise". CivitAI is such a weirdo sometimes. I'll contact support.
@GlowingGuardianGirl CivBot says: "This renaming is a standard part of Civitai's publishing process."
Sooo, not sure if re-uploading will fix this, but I'll try tomorrow!
@cupra I've noticed it happened to other poster's Wan models as well, so I kind of doubt uploading again will fix it.
I'm going to tell the maintainer of LoRA manager on Github, it is already enough when something like ComfyUI breaks things, we don't need sites doing this....
@Mystos Yeah, pretty dumb move for real. No one needs this
@GlowingGuardianGirl I've seen this with every new WAN lora now. They need to fix this, it's annoying :/
@cupra Yeah, they seem to be doing this for about all uploads starting today, I just took a look at some non-video safetensors downloads like Illustrious based loras and it is doing it to them.
@cupra Now trying to use these Loras I just realized how stupid this scheme is, I still use the list style drop down to pick them, so I'm not seeing the "high" and "low" bits on the file name!
Seeesh, it would be nice if the site said why on their update section even if it doesn't have a comment section, but there isn't anything there on this change, and they really don't have a real dedicated forum for people to discuss this. Like I get removing special characters in a file that might be illegal/break on a OS level, but at the same time we also have people using the "_" for spacing since some OS have issues with spaces and I've used Python scripts in the past that don't like Safetensors with periods before the extension.
Might just pose to their dead looking feedback section, don't feel like creating a discord account to see if any site admins are there and actually replie, nor do I want to log into/use Shitter... Plus it still should be made more public.
Oh well the lora manager seems to be working and I closed issue that on Github, looks like it was an issue with Civitai and not the manager, it is now hashing the files and fetching data.
@Mystos It's only working because it's based on the hash, but on the user end it's a freaking disaster.
You can't manage files/downloads properly if their names change.
They even changed the names of files posted before this stoopeed update.
Modifying the name of the OP is infuriating and doesn't make sense.
It can definitely break scripts or everything based on filenames.
They did this without telling.
I've been trying to figure out WHY they did this, but I can't find a single good reason, especially when their system truncates the filename, which can lead to removing precious information and coherence. Users tried to get the filenames right, and they killed it. Some people added the trigger word at the end of the name, now it's gone. Worst part is probably the High/Low for Wan LoRAs...
@Mystos it is so annoying, they have to change this.
I guess it's fixed now :)
File names look like the uploaded ones again.
I feel "Body Expansion" doesn't quite sell it. Maybe call the lora "Thicc Expansion"
Yeah, I'm not happy with the name either. It was the working title and just stuck. I'll try to think of something better. "Thicc Expansion" could work.
Excellent. This has improved so much since you started, very nice work. Really, really great. There have been a few similar LoRAs updated recently, this one is superior by far.
I hope if you make a new update from 2.5 the nipples quality is alot better. Am not feeling it but if you add the K3NK Lora it keeps abit like you said to me previously.
It's because of the low training resolution. Version 3 will include images to fix that, but for now you can always use a low noise LoRA to detail the nipples up.
You can now use this to improve the nipples further: https://civitai.com/models/2311601?modelVersionId=2725000
@cupra Il test out on the weekend
@Devilday666 just need the low noise and reduce the low noise of the expansion lora a little. Good luck!
Could you add height loss/shrinking? Adding "her height increases" to the negative prompt doesn't work.
Unfortunately that was go against the concept of the LoRA, as being an expansion model and also fight the other morphs during training. I'd have to make a separate model for shrinking. Maybe in the future!
@cupra No worries. Really nice work with this btw.
If you push the frame count past the model's training without adjusting the rifle frequency, you will naturally get a reversion towards the original frame. If you're really keen on doing this, that will give you enough frames of shrinkage to encode that motion data and make SVI embeds, which will give a new generation a kickstart in the right generation. 129 frames with freq 0 will get you a good second of shrinkage.
Or you could just play it backwards. That's slightly easier.
Check the ToS before posting though. My shrinkage posts never get through.
2 things noticed: To quote V2.5: "make negative prompts work better" + V2: "try using the TripleKSampler and a negative prompt excluding that effect." Not really, in both cases. You're being too vague. It will mainly require a WanVideoNAG node with an isolated neg prompt (which is not enough either without adjusting the strength of your lora, at least based on my testing).
You're right, the overall concept stays pretty much active all the time. Negative prompts only work to weaken some areas atm, but not remove them completely. We'll see how V3 will do.
@cupra I did achieve the isolation with NAG, but for some reason, the lora affected a male (in same video) instead, and it all just got too messy lol. I decided to go with a Breast Expansion - Wan + M4CROM4STI4 combo (which I should have done in the first place). But I will surely get back to the Body Expansion again at some point (maybe it works well with a DualCFGGuider 🧐).
Is there a premade workflow available for this model? I'm kinda useless in setting these things up but this model looks absolutely amazing so i would love to use it in comfyUI
Unfortunately not yet, but I will release one soon!
@cupra Alright thanks for the response! I'll be waiting patiently :)
Any plans to release an LTX version? Great work BTW!
Yeah, but I don't have audio in my dataset so we'll have to rely on LTX base for that.
extremely good lora, love it. is there a way to minimize face-changing as much as possible?
Reduce the strength maybe or wait for the next release. How is the face changing?
Fantastic, is there a way to prevent pants ripping open with but growth?
If you’re using the TripleKSampler you can try and put it into the negative prompt, along with „nudity“, but other than that I’m afraid it got slightly overtrained in that regard.
Details
Files
wan22_i2v_expansion_v2.5_low_noise.safetensors
Mirrors
wan22_i2v_expansion_v2.5_low_noise.safetensors
wan22_i2v_expansion_v2.5_low_noise.safetensors
wan22I2vExpansionV25.lKBn.safetensors
wan22I2vExpansionV25.lKBn.safetensors
wan22I2vExpansionV25.lKBn.safetensors
wan22I2vExpansionV25.lKBn.safetensors
Body Expansion (Thickness) V2.5 LOW - Wan2.2 - bmorph.safetensors
Body Expansion (Thickness) V2.5 LOW - Wan2.2 - bmorph.safetensors