Finally made a discord server! There's nothing much there yet, and I have no idea what I'm doing, so please bear with me. Just wanted to let you know, since it automatically added everyone in the tip jar tier at least. Let me know if you have any suggestions for channels or anything.
2024-06-26 15:12:20 +0000 UTC
View Post
I think this is pretty close to the final image for panel 2. Just need to inpaint hands for the remaining 4 panels, composite back into a sequence form, and I'll be done. Feedback is welcome!
2024-06-24 14:35:29 +0000 UTC
View Post
I've been working on slightly upscaling and adding details to each image separately. Here is what is currently the final panel although that could change. Next I need to work on face and hand detailing (sounds quick and easy but that's where the majority of the time goes). I think I'm on track to finish sometime next week!
2024-06-21 15:13:38 +0000 UTC
View Post
So I believe this is the raw output I will inpaint the hell out of for this sequence. I'm not completely happy with it, but I love the body language in the first shot, the gradual breast growth, and the half blonde hair in Panel 3. Also, there should be enough room at the top of the frame for dialogue bubbles. I'll have to inpaint Panels 1-2 to be male so that's the biggest challenge here, I think. I'm planning on posting a public tutorial when the sequence is done, but I'll post it here as I...
2024-06-18 14:51:39 +0000 UTC
View Post
So a guy asked to be transformed into Loona, and I thought it would be fun to try animating it! Unfortunately, it turns out that all of the furry models (BB95, SnoutMix, Redwater, and Beast Mix) I tried are too different from base SD1.5 for it to work. I did spend a good amount of time running grids to find the best image settings for BB95 though, so I wanted to share those. I'd highly recommend testing various artists and quality keywords yourself, but this is just a base style I liked.
<...
2024-06-18 04:09:44 +0000 UTC
View Post
Someone on DeviantArt asked me to TG him, so I decided to use the opportunity to test the unsampler node. I was hoping a could get a smoother transition from real photo to AI-generated photo, but it didn't work super well. At a certain point, you still basically have to go full AI, and there's no way to ensure that consistency is maintained. Fortunately, AnimateDiff smooths everything over anyway, so it ended up looking kind of cool! Feedback is welcome!
2024-06-16 04:30:43 +0000 UTC
View Post
Thank you guys for the support this month! Because of it, I was able to rent a GPU for the French maid TG and purchase a better reverb plugin, Much appreciated!
I just wanted to drop a note on what’s next since there are some new people (frankly more than I expected at this point). I’m going to take a week here and do some experimentation and tests. Then I will start working on the receptionist TG/bimbofication sequence with the idea of finishing before the end of June.
Below ...
2024-06-14 04:36:53 +0000 UTC
View Post
We have a tie for which version I should post! I guess that means I get to tie break. Istmael left a comment with the word I'm mainly concerned with: "distract." I think I agree that flashing the trigger word is a distraction, and that's kind of the feedback I've gotten outside Patreon as well. The other version will stay here; it isn't going anywhere if you liked that version, but I'll post the this attached version on DeviantArt. Thanks for voting!
2024-06-13 03:48:25 +0000 UTC
View Post
Which version of the animation should I post?
2024-06-12 01:58:58 +0000 UTC
View Post
It's finally done! Thanks for your patience! So I'm going to use you as test viewers lol. I hope that's okay. Question is, should I flash the word "clean" on screen every time it's said like so? I'm not sure whether it adds or detracts. I'll post a poll, so please vote on which version I should post to DeviantArt tomorrow!
2024-06-12 01:56:37 +0000 UTC
View Post
So close! Just want to tighten up the timings a bit and add sound effects. Should be done tomorrow!
2024-06-11 04:46:36 +0000 UTC
View Post
So I was on track to finish today, but apparently even my 24gb of VRAM are not enough to frame interpolate at 2048x2048. I'm going to have to rent a cloud GPU tomorrow to finish up, but I'm very close. I'd recommend waiting if you don't like spoilers, but I wanted to post something.
2024-06-10 04:34:16 +0000 UTC
View Post
Not 100% happy with this but I think it'll work. There's hair growth before this, but I'll wait for the animation to show that off. I think I'm still on track to finish tomorrow - just need audio basically. I'm really excited to work on sound effects, so if I can't finish those in time, I'll probably just release on here as WIP tomorrow and finish the sound effects before releasing on DeviantArt
2024-06-09 01:48:53 +0000 UTC
View Post
I thought the back view would be easier than the front, but it's turning out to be a nightmare to match the outfit. I like the butt in this one so I think I'm going to just end up inpainting it a shit-ton. Once I get that done, I'll just need to finish the script and audio. Eleven Labs just added the ability to generate sound effects, so if I have time, I'd like to include sound effects as well!
As far as the rest of the month goes, I'm planning a "drawn" style sequence called Th...
2024-06-04 15:00:37 +0000 UTC
View Post
I spent enough time inpainting this that I thought it would be good to release it as a sequence as well as an animation. I'm still shooting to have it done by this weekend, but it's going to be tight. I still need to do the back view which I think will be 4 images in between the images 3-4 of this sequence to show the jean shorts -> thong + butt expansion.
Edit 6/2/24: I got sick on Friday, so I probably won't be done before next weekend unfortunately. Sorry for the d...
2024-05-31 03:36:16 +0000 UTC
View Post
This is a reminder not to fight with the AI. Unless you're a better artist than it is, you will lose lol. One of these images is the one I had in my head for the first shot, and the other is the one I'll probably use. You can probably guess which is which. I tried the DALL-E -> DAZ Studio -> Stable Diffusion process I used for the Latina TG, but I think the image was just too far outside the model's training data in this case. I least now I know what doesn't work!
2024-05-27 02:03:08 +0000 UTC
View Post
So one thing I realized was that I actually hadn't tested whether a morph from English to French accent was possible. Fortunately, it is! I think I need a better male voice model, but I don't hate it! Feedback is welcome. Btw, the sentence is from an interesting site that has a database of this sentence spoken by English speakers from around the world. It's really cool: http://accent.gmu.edu/browse_language.php<...
2024-05-24 04:51:44 +0000 UTC
View Post
Of the tests, I ran this seems to be people's favorite. This used 17gb of VRAM and took 50min to run on a 3090 (15 min to run with LCM but doesn't look as good imo). If you're looking to save on VRAM, I'd just skip the hiresfix part of the workflow. Workflow and needed images attached!
2024-05-23 13:50:01 +0000 UTC
View Post
I think I'm about ready to release the workflow for this. Just need to try it with multiple images. There may be some additional hurdles when I try a more complex animation, but for now it's basically doing what I want it to. Constructive criticism welcome!
2024-05-18 23:36:40 +0000 UTC
View Post
I wasn't planning on posting exclusive stuff here, but I wanted to show this off lol. It's part 3/5 of a sequence which should be done by this weekend. Then it should be full steam ahead on the french maid TG. I'm doing a finetune for that animation that should help with the outfit consistency. It's taking like 16 hours per training run though so slow going.
Edit May 12, 2024: Did I say 16 hours per training run? Sorry, I mean 36 lol. We'll get there eventually. Any...
2024-05-10 02:50:09 +0000 UTC
View Post
Just a test to see if I could transform a character without changing their clothes. I think this is very close and easily fixable with a bit of inpainting.
I want to do another test to see if I can transform a character's clothes without affecting their body, and if it works, I'll post the workflow.
2024-05-01 03:06:19 +0000 UTC
View Post
More tests for the French maid TG thing. I think this method of morphing might be the one I end up using! Next I'm going to try to lock in the characters/outfits better.
2024-04-28 03:42:26 +0000 UTC
View Post
There have been a whole bunch of papers recently with optimizations that have led to new nodes in ComfyUI. I tested a few of them and I'm not that impressed, unfortunately. I used the Limitless Vision v4 checkpoint with the prompt: "a photograph of a woman cleaning a dirty apartment, wearing a french maid outfit, dusting"
A few VERY preliminary observations: PerpNeg and Perturbed Attention Guidance had pretty subtle effects, but increased coherence somewhat. ELLA is very aestheti...
2024-04-27 14:28:57 +0000 UTC
View Post
I've put video testing on hold (need to learn more about still images first) but here is a short test I ran to see if this is even viable. I think it is! I just need to get a lot better at using IPAdapter, controlnets, attention masking, etc. etc.
2024-04-27 14:19:36 +0000 UTC
View Post
I got a little tired of generating maids, so here is something a bit different. The objective was to find the smoothest transition between the subjects. I think my favorite was conditioning average with IPAdapter. I used a male face with face ID to lock the facial features down and then used the "IPAdapter Combine Embeds" node to interpolate between the before and after. Prompt morph is also very good. Let me know what you think!
2024-04-21 14:11:54 +0000 UTC
View Post
While testing Pony Diffusion V6 XL, I discovered that it's actually pretty good at doing sequences! Here is one I tried inpainting, going for a sort of Disney style. I still think there is some style drift unfortunately - I could try either using lower denoising or just paying more attention to that while generating the hires pass. Or just use FluffyRock!
2024-04-07 17:32:19 +0000 UTC
View Post
Just testing the new version of this new checkpoint. It seems quite good at putting men in women's clothes, so I'm seriously considering using it for the French Maid TG!
2024-03-24 23:06:05 +0000 UTC
View Post
Two of the best drawn art checkpoints right now are Fluffyrock Unleashed and Pony Diffusion V6 XL. I decided to do a head-to-head comparison to see which one is best for doing transformation fetish stuff. My conclusion is that Pony Diffusion is better in most cases. Which one do you prefer? If there are any other tests you like me to try, let me know!
2024-03-20 02:58:57 +0000 UTC
View Post
After Effects morph experiment. I'd like to do something similar for the French maid TG!
2024-03-10 15:55:27 +0000 UTC
View Post
Trying to find the best checkpoint for making French maids lol. SDXL seems to be the best, but I'm not sure if ControlNet will work for it. EpicPhotogasm might be as good as it gets...
2024-02-24 21:19:23 +0000 UTC
View Post