I know this thread will pretty much be useless and end up being clutter, but I need help with something and I've not had much help or any responses on other threads. I want to do AI images like the one that's pictured, I'm showing one from we-are-not-ready on deviantart as an example of what I'm trying to achieve.

What AI and prompts would I have to use to generate an image of a woman/celeb who's going through these following scenarios:

Full Body Inflation (Fat/Air, including fat face/puffy cheeks, swollen hands and feet, fully round or round with inflated boobs and ass with stubby limbs).
Breast Expansion
Ass Expansion
Belly Expansion
Turning into a SBBW
Head Inflation (I know that sounds weird, but there were some AI images of that in the Head Inflation thread I liked)
Blueberry Inflation (05 style)
Aunt Marge Style Inflation
Pig TF
Giantess Growth (I know it's an inflation board, but I could mix the two up, like a female hot air balloon)
Legs Growing Longer (Again, I know it's weird)
Muscle Growth
Foot Growth

Also with clothes ripping, shoe busting, buttons popping or full nude?

I know if seems like I'm asking a lot, but if someone can help me with this, I'll be very grateful and we can let this thread die afterwards.

>>105284 (OP)
While it's a good start there's an ai board for some reason and we don't know why we have so many boards.

.

.

.

.

.

First odd, there's a good AI thread in another board on the site that's got lots of details on various AI techniques:
https://bbw-chan.link/bbwai/

Now, it's true that the above thread is more generalized toward BBW/SSBBW AI generation, but all the basic the technical stuff all still applies.

It comes down to choosing a model that's been trained on the appropriate content to begin with; same goes for LORAs and other additional tools. And engineering your prompts based on what works best for the model and tools you're using is essential.

A lot of the commercial AIs have some form of censoring built-in, which can make it difficult to produce partial or full nudes. Inflation content is possible as the models pick up more terms and tags that can generate it -- nobody seems to be pruning that our, at least.

For generating WG, inflation, hyperpreg and everything in between, Stable Diffusion running the Bigger Girls V2 model can work well. You can try it out on stuffer(dot)ai for free. If you really want to pursue it, it's better to install Stable Diffusion on your own computer, and set up the models and tools the way you want -- then since you're not waiting on a shared system, you can try out and learn various approaches to prompting and SD settings.

Depending on how an AI-generated image was saved as well as how it was affected by post-processing/editing, many times the prompt is still embedded in the EXIF metadata that's embedded in the image file. Stable Diffusion can read and display it directly in the SD application. And there are prompt reader/extractor tools which may be able to pull it out. It's a good way to see how prompts are put together and learn prompt engineering -- as long as the metadata is still there in the image.

We don't have an inflation-focused AI image thread currently, so it's fine if this one can survive. One or two others saged into oblivion because they got pushed down the list by too many clones of other threads. It might be nice to let this one live.

Pic related. I hack around with Stable Diffusion sometimes. I'm really a writer though. But kicking out AI images sometimes is good for a story idea and it's fun. I post on DA as "joemonday".

>>105331

I have looked in there, a lot of it is very confusing to me. Is there an app version of this Stable Diffusion I could use? I tried that stuffer one, can't quite get the hang of it.

>>105361
The "automatic1111 webui" (I suggest googling for it; it's on Github) version of Stable Diffusion is the easiest to set up. There's even a Windows installer that will drop the necessary Python files and configuration. From there, you launch it locally in your web browser.

Setting up models, dependencies, and figuring out settings once they're working is the hard part. Not all the models and dependencies are documented all that well, so it takes some trial and error. The AI board on BBW-Chan has plenty of ongoing threads about that, and you can get ideas from there at least. Sometimes it just takes a lot of web searching to find hints for stuff. But it's fun once it gets going.

stufferai displays the prompts with the images on the webpage, so if you see something you like, you can see the prompt and use that to figure out what might work for prompting.

And the Bigger Girls V2 model uses danbooru tags in its training, so using tags you want in the prompt helps a lot.

We shouldn't have an AI board if there was a rule to indiscriminately report AI content outside of one single AI containment general.

>>106397
I would be fine with ONE thread for AI inflation. Probably better than a single AI thread lumping together BBW, SSBBW, Immobile, BE and everything else. Nobody on a chan seems to be able to ignore what doesn't interest them.

So an AI inflation thread -- or maybe AI realistic and AI cartoon style pair of threads. BUT ONLY ONE THREAD PER TYPE DAMMIT!

I don't make decisions around here though. Above my pay grade.

>>106776
This. This so much.

So I peruse stufferai once in a while. seems like the "BBW Aurora Borealis" model can do some decent bloating/inflation. The prompts get listed with the images, in this case whoever did the prompt included "small breasts" and "small chest" in the prompt to get the look of the character's torso just ballooning up as a whole which is kind of interesting. At least AI sometimes can do other than the typical inflation shapes. (And sometimes it just screws it all up in strange ways...) Unfortunately AI still sucks at hands.

>>105284 (OP)
Do you think you can be active on the superfluous AI board?

>>105284 (OP)
See >>106840 (Cross-thread)

I tried making some in SD with PonyXL. There's some promise here, someone with better skill than me can probably do it better.
WARNING: Gore (a little)

>>109385
How? More?

>>109447
I made a few more, but none of them were as good (some even quite bad).
If you wanna give it a shot, maybe even improve it, I can share the parameters:

score_9, score_8_up, score_7_up, score_6_up, 1girl, female, solo, closeup of a woman strapped to a metal chair, leaning back, hose connected to her mouth, inflation, very puffy cheeks, impossibly large cheeks, inflated cheeks, closed eyes, tears of pain, in pain, throat bulge, blood, veins, veiny, veiny belly, sweat, medium dark hair, in a dark basement, dark basement background, source_anime, rating_explicit
Negative prompt: simple background, angry, signature
Steps: 35, Sampler: Euler a, CFG scale: 9, Seed: 1368155992, Size: 1024x1024, Model hash: 67ab2fd8ec, Model: ponyDiffusionV6XL_v6StartWithThisOne, Clip skip: 2, Version: v1.7.0

>>109508
The face is fucked and the red water looks like blood and yet doesn't. And post more.

>>109525
That's all. The rest are legitimately fucked. Eh, I'll post one of the failures for some laughs.

So are we allowed to post AI on this board now? The dedicated AI board is kinda dead. Seems like the people who hate AI could easily avoid a single thread if we keep it contained. Who makes the decision?

>>109614
Go for it, its a better topic than 90% of the hyper-specific trash threads that clog this place up. Also, your stories/scenarios are great! Would love to see more non-berry full-body stuff from you.

I'm glad we have this here since in bbwai the fat crowd hijacks the inflation only board

>>109624
Thanks! I'll post some others here for those who don't have a DA and can't see them otherwise.
My stuff has definitely been blueberry heavy lately. That is what most watchers follow me for, but I try to do like 1/3 non-blueberry inflations. It's actually a relief to do those because they're maybe half as difficult as blueberry.
cheeseAI is a great follow BTW if you want non-blueberry.

>>109614
>>109628
AI should've never gotten its own board regardless of how loud or how many autists were demanding. More boards mean more hassle to mod the damn thing.

>>109770
Good.

Keep the pressure on them never let up

>>109635
>>109635
>>109635
waaah! what incredible work! Can you explain to us how you managed to obtain such a clean result? What AI did you use? what models did you use?
Thanks a lot

>>109771
No it's not good. Giving in to the demand of autists is never goodl.

>>109775
The art itself is pretty mid as far as AI goes. Theres alot of imperfection. The models they are using clearly are meant for facial expressions, and the bodies end up looking wrong.

The anatomy is poor, and im sure it would have looked even worse if it wasnt tidied up in photoshop.

Its almost certainly Stable Diffusion, running some bog standard Japanese girl trained model, with the Hyperfusion LORA for any semblance of inflated shapes.

Its lacking controlnets, and better LORAs and prompting techniques to give better shape, form, and prevent or mitigate concept-bleed. The only thing these images have going for them is the resolution and faces, which is incredibly easy to do with minimal setup.

So don't be amazed, its bottom of the barrel. Atleast without being a reject AI blooper.

>>109775
Thanks! As the other anon pointed out, it is Stable Diffusion with the Hyperfusion LoRA. I use the RealCartoon-Realistic checkpoint.

>>109791
You are absolutely right that I'm not using any controlnets and so forth. What would you recommend as a better trained LoRA for inflated shapes? You seem to know a lot more about this than I do. I'd definitely love to see some examples of your work as well to see what you're talking about.

>generate a pregnant babe
>generate a hose and a pump
>???
>profit
Profit.

You absolutely need to try the AI generals on 4chan. They're great.

Can you turn this into an AI-generated image?

Not so great....

>>111305
Do something about your small penis becUse I welcome it. With limitations lol

>>111305
The first is serviceable. The second is... interesting, at least

>>111305
Good attempt. Let's wait until they're better. Next question: Can you turn a random image into waifu fanart? For example, if I give you the image of >111235 and told you Peach or Samus, could you do it?

>>111387
You jumped in your suv to type that? Pathetic.

>>105284 (OP)
I was thinking the other day... could you imagine if someone fed a bunch of TMC videos to an AI? I wonder what video it would spew out?

>>
Feds coming for you like nycha

>>113054
Who's Nycha?

>>113055
70 employees arrested for corrutpion & hush money secrets for time off to other groups

>>113056
womp womp

I have a question on how to make ai inflation generate pics?

Sometimes pregnancy and even fat gens look more like inflation.
Pics related are by ga1nzz.

>>113149
>pics didn't attach
I'll see why's that and post them later.

>>113054

Who were you talking to?

>>113179
Long island

>>113149

>>113200
Now u see why your brother bashed his family in high school. Seeing it for yourself now.

>>113202
After the tax return money stop coming in noticed how they treat you like shit. Dont worry all that money is coming back to you. Theyre not holy.

>>113203
>>113203
They had you to make money off you. Look through the years. They never praised you. Bo birthdays no nothing. Kept the money to themselves. Thats not love. God got something 10x better in store for youX

Your god is not real. You treat your kids like shit. You lack health when you treat your kids like shit. Pure hearted age good and so does the devil

i think there is an ai called VIGGLE wich makes a person on a video to be replaced with another, maybe this can work for replace actors in inflation scenes?

>>105284 (OP)
https://storydiffusion.github.io/

>>113578
Its crazy making over $200 per hour offends people. The uncomfortable conversations I like to have.

I request the first image and got the second one on 4chan.

>this get
Wew.

>>113666
Is there a free image-to-image AI that lets you request adult content like this one?

>>105284 (OP)
For head, I'd advise using Dall E 3. The best prompt I've found is

a woman whose big head is inflating round like a balloon with a bigger head

Redundancy works.

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

>>114626
Oh yeah she has a death wish. Her kids are on our radar.

>>114576
I love the women with the hose up their ass. It’s a bidet inflation fetish dreamboat.

>>105284 (OP)
I'm learning stable diffusion but I wanna add the face of a celebrity to some inflation images. I just cant get it right. I highlight the face and it just pastes it in the art, but not in the right place. Got a lot to learn. Anyone had any luck with putting a celeb face on an AI generated inflated body? I just want scarlet johanson with a huge inflated butt. I've seen some, but I want to make my own. Thanks.

>>114626
Dude! I love RE. I'd love to make something like this. Any tips?

I use Stable Diffusion

A good use of AI is photomanipulations. If you can roughly edit a morph then AI can be more informed by using InPaint to generate something with more true to image quality. It unfortunately loses that loot box fun, but if used correctly you get solid results.

>>114739
Can you make her fatter and dirtier? Like a back porch 600lb life sorta affair.

>>114753
Move back to new york bro. You dont need thT.

https://www.deviantart.com/carlitosmurio

Let's share prompts and methods to reverse-engineer them and find out how to create excellence.
I'll start:
(score_9:1.3), (score_8_up:1.2), (score_7_up:1.1), source_anime BREAK
1girl, may \(pokemon\), outdoors, forest, foliage, squatting, hose, anal object insertion, (pregnant:0.8), smile, blush, spoken heart, pussy, twitching, heavy breathing, drooling, completely nude, female masturbation, fingering, orgasm
<lora:sdxl_yd_v1:0.5> <lora:sdxl_trogfor_v1:0.5>

https://www.deviantart.com/carlitosmurio

More inflated anime chicks with AI tools please!

I like a lot of these anime girl generations, but the sameface is getting a little... well... same-y.

I know a lot of the SD models based on the BiggerGirls models have kind of a problem with it, but you can try different things in the prompt in different runs to try for other results. It should be able to recognize "ahegao" along with others to modify the face, like "half smile", "smile", "happy", "mouth open", "bored" etc. for better or for worse, some expression cues may also affect body postures too.

Mess aroudn with it. Pull the slot machine handle and see what it spits out.

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

moar belly inflated anime chicks via air enema plz

https://www.deviantart.com/carlitosmurio

Have some real "bangers"

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

Hi !
I'm trying to create huge breasts girls
I'm a begginner

https://www.deviantart.com/blaalbblaalbblaalb2

>>117967
These are great! I'm new too. What methods did you use? I have some inflation loras but can't seem to make anything good. I just want to make people look inflated, not just have big boobs. These look like blown up boobs. Awesome

https://web.archive.org/web/20240623004654/https://pregchan.com/d/res/32336.html

You need to be culturally appropriating the pregfags.

>>118133
How did you get that second image?

>>118130
I put a Samus in a jar and farted until I shit

>>118183
I found it on Pregchan.

https://www.deviantart.com/carlitosmurio

A mate sent me this.

POSITIVE
score_9, score_8_up, score_7_up, score_6_up, anime_source
hyuuga_hinata, naruto_(series), fishnet
sexy,
belly expansion , fishnet bodysuit, garden hose in her ass, anal object insertion, water,
pain, blood from mouth, nosebleed, bleeding, blood

NEGATIVE:
3D, CGI, render, photo, text, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, distorted, blurry, glossy, blur, jpeg artifacts, compression artifacts, low-resolution, bad, grainy, symmetrical, duplicate, error, pattern, beginner, pixelated, fake, hyper, glitch, overexposed, high-contrast, bad-contrast,
censored, mosaic censoring, bar censor, blurry, bloom, head out of frame, watermark, worst quality, low quality, jpeg artifacts,

With the InflationXL-02-PonyXL-e050 Lora at weight 1
I use comfyUI

I haven't tried it and frankly I'm a bit afraid to.

https://www.deviantart.com/carlitosmurio

https://www.deviantart.com/carlitosmurio

Can anyone provide links to the best ai generators that are free?

>>119330
stuffer(dot)ai
Uses Stable Diffusion, configured with the most popular BBW models and several are also good for inflation. You just have to work out the prompts.

>>119501
Some things I've gotten out of it

Requesting one or all of these 2d-ified.

https://www.deviantart.com/carlitosmurio

Some folks out there making neat stuff with video

>>119848
God damn this is good. Source?

>>119848
We NEED a tutorial for this!
Even just the image and not the video. I can never get the mouths and cheeks like that, also, the boobs look amazing. Teach me wise one.

>>119851
Learn to read filenames.
https://www.deviantart.com/kitteninflator/art/KPop-girls-trying-not-to-actually-pop-1072216166

hey, someone knows how to make inflation videos in Luma ai? i tried making some prompts but all of them fail

https://www.deviantart.com/carlitosmurio

hey, actually i made up some videos on Luma and i realized, is not that hard, its even surprisingly underwhelming.
Just make an image for the inflatee, deflated, then use an image editor with ai of your preference, put the two images on luma and write down "woman blowing up like a balloon" and voalá

i made some other videos using hellresident's sequences...and some strange stuff happened......but i think they work atleast

>>120171
Honestly need more of these keep them coming and share you're processes I believe someone hellresident stuff just got posted in lost media maybe try that?

>>120176
Thanks, but the daily limit of Luma ai makes it difficult for me, my intentions were trying to show that is possible to make AI inflation videos with luma
:3

>>120119
I need the prompt for your image #4.

>>120205
We honestly need all the prompts if you won't continue the legacy at least allow us to pick up the sword and slay the blimp dragon for you by overinflating it till it pops with the legendary air pump!

Another one by kitteninflator

i tried to make some videos more closer to my kinks, but looks pretty off, i tried this like 5 times, the others 4 looks awfull, this one atleast does what it was meant to do. But well, is buff woman inflation

>>120364
More buff woman rapid weight gain

AI becomes more fappable as the time goes by. Awaiting realistic AI generated videos in upcoming couple years

>>120478
Really happy for AI progression and getting better. I am sick of artists bitching about lack of free time or motivation. Soon they won't be needed and you will be able to generate anything you want in any artists style paying around 30 bucks a month, instead of greedy 150 dollar commissions just for 1 image, and that is if you manage to outbid others for the spot.

>>120481
Which is hilarious to say given how many hundreds of AI Patreon accounts there are now.

>>120496
Yeah, because people try to profit from others not knowing how to make good AI inflation art, that still requiresto write codes, filters and knowinghow to manipulate the tool. When it will be even more accessible to broad public, when it will be enough to just write plain text of your fantasy and feed some images of your favorite artist and your done - tons of fappable images merely for free. Thar way all those Patreons will die out same way the artists will.

>>120548
And, of course, assuming the computers don't get borked like a couple days ago. lol

>>120478
First two look really nice. Can you provide the source?

I've stumbled on a really good ai generator called Hi AI, it's one of the Talkie clones and it's uncensored, the keyword to get is pregnant and you can get the character very inflated.

The app is available on Android might be for apple

https://www.deviantart.com/carlitosmurio

Has anyone made anything good with novelai’s image gen? Messed around with it and I like how it looks, but I’m struggling to make good full body pics.

I wonder if AI can turn inflation stories into artwork/comic?

>>121194
Maybe someday. But not yet.

What you CAN do is take a good description from a story and condense or edit it down into a prompt that the AI can parse.

But it's still not going to work incredibly well unless you use keywords that the model was trained on, group specific relevant ones together with parenthesis for emphasis, and also feed it a properly constructed negative prompt to filter out unwanted AI generation behavior and keep the thing on-track.

So copy-paste from a story won't work. (Not yet, at least.). But you can sort of boil a written description down into something AI can use, once you get used to how the AI and the model interpret stuff.

>>121220
prompts and loras? mine never look this good lol

https://www.deviantart.com/carlitosmurio

I'm only just getting started with the XL models and LoRA, but I don't see many other people presenting their tools & recipes so here's mine FWIW.

Checkpoint:
Pony Realism
>https://civitai.com/models/372465/pony-realism

LoRA:
Inflation XL
>https://civitai.com/models/232970?modelVersionId=569129
use the terms described in the SD version:
>https://civitai.com/models/210476/inflation-lora

Also helpful:
Hands XL LoRA
>https://civitai.com/models/200255/hands-xl-sd-15
for better hands

Positive Prompt:
read the description for the Pony Realism checkpoint, there are some terms you should include:
>score_9, score_8_up, score_7_up, BREAK
I place these toward the end of the positive prompt, just before the LoRAs
I place the LoRAs toward the end
I'm using "spherical inflation" at the max recommended emphasis of 1.5
A slightly wide frame can produce a rounder body, like 1152x1024, or 1024x960
I'm using the inflation LoRA at an emphasis of 0.6 like so
><lora:InflationXL-03-Pony-e010:0.6>
Also the hands:
><lora:hand 4:0.6>

Negative prompt:
From the checkpoint description, I place these toward the beginning:
>score_4, score_5, score_6
I also tend to place terms like "fat rolls" and "obese" in the negative prompt because that's not what I'm going for. I want tight & inflated girls.

None of this is definitive and I've only just started. If others have better info, please chime in.

>>121646
Thanks for the info! I like what you've coaxed it into producing.

I hack around with SD and BiggerGirls model B v2. It can turn out some acceptable, fun results with enough prompt tweaking, but other models and LoRAs are really moving beyond it.

How is it for time to generate images? At least with my current setup, generation is fast, which makes it easier to get results and then play around with the prompting to get it to kick out useful results.

>>121893
Thanks. "Coaxed" is exactly the right verb to use with AI.
For me, it could be faster to generate a ton at low res and then go back and run the ones I liked at higher resolution, but I find that tedious. The best method for me is to generate at full res with "generate forever" on, and skip the ones that are not looking good enough.
Although you do have to be careful about cancelling too soon because sometimes the images improve quite a bit in the last steps.
My setup is reasonably quick, what takes me a long time is editing and refining, and coming up with captions and sequences.
Not sure what tools you're using but I always say that the negative prompts are the most neglected. Think of the opposite of what you want and type that in there.

>>121900
Sounds like your workflow is pretty similar to how I prod SD. And I've noticed the same thing, even with a different model. It will kick out image after image that's not OK for one reason or another, and then just when you think there's something wrong with your prompt it will start spitting out workable images. It's just weird how it behaves.

Likewise, I also found that negative prompts can go a long way to keeping it "on track" with the intent of the main prompt. That, and depending on the model, figuring out by trial and error if parts of the prompt which should be directing it toward what you want are actually confusing or misdirecting it. It's not like programming where very specific syntax will produce expected results, or at least traceable errors that can be resolved. SD doesn't have a mind of its own but it sure has a tendency to veer off into the weeds if left to its own devices.

Once I have decent images, I'll retouch if necessary with Paint.net or GIMP. Sometimes I'll try inpainting, but it's always been a little hit or miss for me. I once spent an hour trying to coax it into fixing a foot. One freakin' foot. Geez.

I tend to prefer to leave the picture un-captioned, and drop captioning into the description text on DA. That way folks can read it if they want or make of the pic what they want on their own. That's just me, though. I started out a long time ago writing some inflation stories, and that can take a little while. SD generation lets me condense general scene ideas into a prompt to (hopefully) get a pic related. Or sometimes it kicks out something cool, and then I wind up sitting on on it, trying to come up with the perfect few lines for a scene related to the image. I have a backlog of enough to keep putting out a few per month on DA and not have to generate any more for a year. But of course I keep going back and poking at it because I might be able to get it to do something interesting and new.

I have to figure out if I can just clone my install of Automatic 1111 and run a separate instance in another folder with a different model and settings. Sure you can import and select different models in the webui, but there's no great way to back up all of the settings that go along with a model as a preset or anything that I've found. I've got it behaving pretty well with the model I'm using and I don't want to lose a successful setup.

guys i found an AI wich makes videos faster and better than Luma, atleast under my opinion,
its called Vidu
https://www.vidu.studio/create

i used prompts that were like "action movie. a woman, her whole body bloats and blows up like a gigantic and obese human water balloon woman, her slender shape turns into a bloated and rounder one, her belly expands like a balloon being blown up"

>>122063
sadly it triggers a warning sometimes, when that happens or i slighty change the prompt, or i keep generating,

I tried to make a video using it, thanks to the warning it was hard, and it doesn't read my prompts very well. But on other side the inflation looks really good, literally made a decent inflation video with a single prompt, also it makes it freaking bouncy wtf! I tried to make a video of jasper inflating, wich made her look weird, but the inflation looks very nice. Remember to use words like water balloon, i think that makes it jiggly and bouncy

It's does nudity without out asking sometimes
https://www.vidu.studio/share/2382954625897570/516394

Depends on the situation

i also made this
https://www.vidu.studio/share/2383379906458215/226256

this was the prompt

i think the Ai hates me, or my prompts are too obvious to be nsfw
"action movie, on a forest. a very pretty, female bodybuilder, her whole body bloats and blows up like a gigantic and obese human water balloon woman, her fit shape turns into a bloated and rounder one, her belly expands like a huge water balloon being blown up"

>>122093
>>122103
Damn these are really good

This one came with a story, in the start the girl dies suddenly and In her past timeline she's wondering around an abandoned warehouse ang she gets inflated and popped causing her to die in the present day
https://www.vidu.studio/share/2383250752298375/088387

>>122063
What is up with luma anyway? ive been in queue for one video for the last 2 days.

>>122063
Thats because Vidu is using the actual SORA architecture, from the impressive proof of concept videos a few months back.

>>122408
Interesting. But this means the ai will "corrupt" into a non-inflatuinship version? Bc i noticed that by the pass of time the results look slighty worse, and the trigger is easier to activate. Some tips?

https://www.deviantart.com/carlitosmurio

THE FUTURE IS HERE
flux1.org

Oh badass

>>122557
I love the enthusiasm but I've been getting more errors then generations and it punishes you for making an error and times you out for 60 seconds every failed attempt, what exactly is your prompt?

>>122555
Awesome. What were your prompts?

ok, this thing can make good videos if it wants to

here's something I tried with EbSynth

>>122078
HOLY SHIT this is next level. Its too bad its still finicky or else this would be a godsent. Still hope people post what they can from it here in this thread.

>>122671
thanks, also i tried to re-use the prompt, but for a strange reason the results dont go as this one, or maybe is bc i haven't tried enought? btw this is the prompt, maybe someone can "fix it" for make it even better. But if we manage to control this AI, we can make our inflation scenes
"action movie, on a beach. a beautiful female weightlifter, her whole body bloats and blows up like a gigantic and obese human water balloon woman, her fit shape turns into a bloated and rounder one, her round body expand like a gigantic waterballoon being blown up"

>>122660
Lazy git, don't gives us bull, that's from bendybrooks X feed

Bad news guys the Devs have changed it and put on a strict filter

I found a good alternative https://magicstudio.com/

Someone knows a way to evade Vidu's triggerings, or an ai similar to Vidu?

Running SD locally is still working well for me

vidu not bad

>>122756
Put that in 0.25 and it's pretty good. Maybe slowing it down and then interpolating frames would be perfect

>>122756
what prompt did you use?

Not too bad.

>>122709
Nothing gets past you, does it, genius?
I tried making it more life-like with img2img. Took a lot of tries to get it right, and then I had to photoshop the relevant portion into the original frame, so, if anything, I went the extra mile.
You're welcome.

someone knows any good video2video ai?

I made these some months ago but didn't bother posting as they're not much to look at. There's no real expansion among the Violet characters and the blueberry girls are just sort of walking or standing.
These are the sort of results you can get with AnimateDiff. Just sort of characters standing around. Fun enough for me but maybe for others it's just meh.
Posting in case anyone likes.

>>122751
sauce?

>>122810
It's ai there's no sauce

>>122771
Good shit dude. Prompts?

>>120171
how you type what type?

how'd I do?

>>122892
Ok, we must be using different Vidus cuz not Vidu is just releasing glitches. What are your prompts?

>>122892
In other words, youre doing amazing, keep going

>>122890
What?

>>122895
Just using standard vidu with lots of trial and error, it all comes down to the starting image an a lot of luck.
My prompts were: “a woman inflates like a water balloon starting from her belly”
And for the blueberry: “a woman turning blue with the colour starting at her nose then spreading, once she is fully blue her body widens and her belly inflates like a water balloon”

Has flux got censered?

>>122900
sorry i ask what you type to get video so good like blueberry is amazing

>>123001
Thanks, thr prompt actually is pretty basic, i dont remember well, something like.
"Horror movie, a woman swelling up tremendously like a balloon, her skin turns blue" or smt like that. The thing is that i used luma's Image feature that allows me to add the starting and final frame, so i used a sequence by Hellresident and post it there. Sadly the seek for sequences that both are live action and look like the same person is hard, and the loading time of Luma is horrible. So i opted for other optionss, like Vidu wich i need to find a way to surprass the freaking warnings wich i would like some help

>>123050
ok, how do you skip the triggering thing? what methods do you use? Everytime i try to make a video, or Vidu makes it wrong, or it shows the warning

>>123009
thank you! you do not have history anymore? i try prompt but it is not good like you

>>123052
Theres two triggers. Before prompt and after prompt.
Before prompt requires rewording. Avoid words like "thick".
After prompt appears to be a filter to remove already generated images that "conflict" against policy. Your only solution is to regen. But depending on how risque your prompt it, it may take a while.

>>123070
So the method is doing it over and iver? Ok

>>123066
You need the images or else it wouldnt work

>>123098
tyty!>>123098

How useful is the "Enhance prompt" check on Luma? Does it do anything?

i wonder if someday i will be able to remake this vid. Maybe with vidu? but i would need some help

>>113498
mmm, i haven't thought on that idea

>>123303
Btw what are your prompts, it would maybe work on Vidu

>>105284 (OP)
First of we need a prompt anchor.
>>105285
I'll treat this as a prompt anchor. Here's one I got from asking.

everything in the curly brackets is a dynamic prompt and XXXX are wildcards:
>pos
score_9, score_8_up, score_7_up, source_anime,uncensored,BREAK,may (pokemon),1girl,solo,completely nude,BREAK,huge breasts, nipples,{huge pregnant belly|pregnant|belly},navel,BREAK,(wide hips,huge ass:1.2),BREAK,anus,(pubic hair,anal hair:1.2),BREAK,erect clitoris,pussy,anus,(pussy juice:0.8),BREAK,blue eyes,looking {at viewer|back,ass,anus} ,{pos_emotion|neg_emotion|ahegao,drool},blush,BREAK,brown hair,BREAK,poses,anal hose insertion,fingernails,BREAK,colors gradient background,
>neg
source_pony,source_furry,source_cartoon,censored,(hands:1.3),long fingernails,watermark,logo,artist name,double dildo,gaping,shower,lactation,stray pubic hair,cables,tan,loli,flat chest,long legs,tall female,holding,disembodied,handjob, limb,faceless male,double penetration,monochrome

The prompt is not ideal and still kinda gacha, but it was good enough for me after all.Post a gen if you can, i'm curious to see if this prompt works for you aswell! I also boxed the wildcard files so you can check their contents too.
And don't forget that you can always take any gen from the anal hose mega and get the metadata from there too.
>neg_emotion.txt https://files.catbox.moe/vezx9z.txt
>pos_emotion.txt https://files.catbox.moe/0ol3a7.txt
>poses.txt https://files.catbox.moe/wpr4p2.txt
>colors.txt https://files.catbox.moe/fltl72.txt
I don't use a concept lora for the anal hose. And for the style i mix my cromachina lora at 0.25 with my takku lora at 0.35 strength.

>>123399
Only good for pony, which means any realistic results are out the window. Parts of the model will always be stylized or unrealistic regardless of merge or LORA to attempt to fix it.

For photorealistic images, I have a starting prompt of "Full body shot by a Nikon D7 of a..."
The full body shot usually ensures the whole body is in frame. And the Nikon D7 ensures photorealistic qualities. You can expect prompt drift with that however, as it tends to start embedding those words into the image. Especially with complex images with simulations of text. This can also be interchanged with Stock Photo or 35mm Film.

People are also wildly uncreative and too dependant on LORAs, which restrict what your prompt can do. Whatever your LORA was trained on, will start to make your images samey.

You can achieve a lot of unrealistic proportions with keywords such as "Photoshopped" or "Photomanipulated" to help the AI search for altered images, and expand proportions in a clean way without ugly skin artifacts.

Hey vimu can work if it wants, too bad it turned the belly into a balloon rolling over her head after. So I cut it down and put some sound over it.

>>122103

I also went and added some sound to that one.

>>123645
>>123650
Great vids, also how you make the sound? You add sfxs using a normal video editor or Vidu (or some other AI) makes it for you?.
Also can you please make sfx for this vid too?
>>122078
>>122639

when you make videos on vidu you let the "enchance prompt" checked or unchecked?

by cheeseai on deviantart

Here, op. Fellow weirdo. I'm gradually figuring out how to get Luma to do head inflation.

The first one's prompt is "A woman's head slowly swells up and inflates tremendously". The second's is "A woman's head expands and inflates tremendously".

If someone has a better way to describe it than these I'd like to hear it.

>>123908
Also ofc this is using the start and end frame function

>>123908
Btw Luma's waiting times are getting better or worse?
They look good, not my cup tea but thats ok, i would recommend to makr their faces look puffier, like puffy cheeks, etc

>>123914
Also i would like to find an Ai to make sequences, but i cant find one, cuz morphing an image normally makes an static video.

someone knows a way to make good looking inflation vids at RunwayMl, i mean cuz now they have Gen-3 wich looks awesome, but doesn't make great vids like Vidu (when it wants to make)

https://www.deviantart.com/carlitosmurio

i tried to use vigge to make some vids, It didn't end up very well....but at my generations i saw a strange..patron..the character actually streetches, we just need to find, at what kind of circustances does. This video its technically glitched and silly, even disturbing, but its a step to see if we can make something better
Here i have an edited clip, the two clips inedited, the 2 sources and the image i used

i tried to post all of this on a single post but this site didnt want to, letting me only post the video and this image. The sources are RIPD and the school nurse files

mmm, its slighty to become better

>>124117
>>124132
Not good/10

>>124149
i know, thats of what i was talking about, the thing is we must find a way to make it work, cuz for a strange reason i fell like there is an strange patron in here...... OOoor look if there is another ai like this but better. Btw what happened to the Video2Video? the last one i heard of was Morph Studio

I've been having fun with Flux it's way better then grok, grok crashes after a few generations, compared to dalle3 flux censorship is bare minimum

>>124117
Damn its a shame Vigge is still so rough. Been DYING to find a good female version of the Fat Elvis scene from RIPD. This is the closest so far unfortunately.

>>124326
i mean, i think video2video can also work, the issue is that i dont know any free on web video2video ai that makes high quality videos

>>124326
Also yeah, an strongfat monster girl would be perfect ;-;

I saw this guy appear on deviantart who managed to get big round butts and small thighs down to a science.

https://www.deviantart.com/expandedyummyness/gallery

I haven't been able to really replicate this on my end with stable diffusion, so any tips with the models or prompts would be much appreciated. It's hard to isolate exactly what gives you this result. "Bubble butt" will sometimes generate literal bubbles, "mega colossal ass" will generate some flab, and negative prompts such as "thick thighs" either won't work or will override the big butt.

I found an Ai called Ltx Studio wich supossely makes full lenght videos, also i think has a keyframe video generator. Someone can see if can make content?

there is another video generator ai called Hailuoai, maybe it can make better videos?
https://hailuoai.com/video

>>124523 ExpandedYumminess uses a Flux model. If you use sd-forge locally, you can see the generation info he uses in the PNG info tab. Everything is there, even the Lora hashes, so you can get them from civitai. Trying out various values for the "Diffusion in Low Bits" field, I could exactly reproduce his result.

The attached image is a repdouction, not a copy.

>>124599
I can't get anything. Everything I download is in jpg format, so no data for me.

CivitAI? never used it before.

>>124666
It does not matter, whether the image is a jpg or png for the analyzer. For the image in question, it is a jpg file, and the resulting generation data was

big tongue woman, pigtails, at home, dinner table, pizza on plate, tight shirt, big boobs
Negative prompt:
Steps: 30, Sampler: Undefined, CFG scale: 3.5, Seed: 578216591, Size: 832x1216, Clip skip: 2, Created Date: 2024-08-27T05:40:58.9714362Z, Civitai resources: [{"type":"checkpoint","modelVersionId":691639,"modelName":"FLUX","modelVersionName":"Dev"},{"type":"lora","weight":1,"modelVersionId":739724,"modelName":"Perfect Tongue Mommy","modelVersionName":"V1"},{"type":"lora","weight":1,"modelVersionId":734465,"modelName":"Big Boobs FLUX","modelVersionName":"V1"}], Civitai metadata: {"remixOfId":25014290}


In order to perfectly reproduce was ExpandedYumminess is doing, you have to reproduce their setup. There are also models by ExpandedYumminess on civitai:
https://civitai.com/user/expandedyummyness

Good luck.

As an amusing note… If you use automatic face detection with adetailer, it often thinks that the blueberry's nipples and belly button are a face on a giant blue head.

>>124718
Thanks. I hate it

>>124718
When I've tried it, adetailer gets easily confused with any massive expansion. I've had better luck masking and re-generating faces on otherwise good outputs. I'd love it if it were reliable with expansion models though. Would save some time and hassle.

Hilariously, probably in order to avoid inflation because it's a kink, some AIs will accidentally turn inflation into a different kind of inflation. I tried to make body inflation and ended up with breast inflation instead

btw someone found a better text to video ai for make inflation videos?

i think i decode it. on Hailouai we must use this prompt
fantasy movie, a woman drinking a glass of water, her body bloats and blows up into a huge extremely obese body
it worked, i dont know how. Oh! and also desactivate the better prompt thing, that ruins it

i used Leonardo AI to make the both images and Kling AI to make the video and oh, wow. I think we are cooking
"Science fiction movie, A woman blowing up like a huge balloon" 0.6 creative

here are some other vids made by Kling ai, the lore is that she is as soldier woman on an testing facility who get infected with a virus that causes her to inflate

Vidu works ok. Good for some belly work on some genderbent pyrocynical ai images.

closer attempt to make inflation art using Hailoau ai
"2000's horror movie, a beautiful woman wearing a swimsuit, she swells and bloats and bulges into a huge inflated obese woman"

https://www.deviantart.com/carlitosmurio

Idea for anyone else who does a lot of inpainting. If you make some margins, you can draw bursting clothes or milk or whatever seeming to come out of the frame with a trompe l'oeil effect. Makes for a more dynamic image.

this is the best i could get with Hailuoa
"2024 horror movie, dramatic, a beautiful woman wearing a swimsuit, she blows up and bloats into an huge obese inflated woman, full transformation on screen"

another attemps made by kling ai and prome ai
"science fiction movie, a beautiful woman wearing a cyan swimsuit, her body bloats and swells grotesquely like a huge human waterballoon being filled with water" creativity-revelance 0.6

>>125448
how are you making your ai art? i cant seem to get anything to look anywhere nearly as good