DuxVector
patreon
July 2025 Monthly Report, V4.6 visual update, new poses and more!
🕑 Added 2025-08-01 20:01:01 +0000 UTCLuci-chan quick!!! You have to try out the new update potion! The potion seller said it's too powerful for me, but you're a demon girl so you should be able to handle it!
TL;DR (SHORT VERSION) -
V4.6 update. Massive increase in visual fidelity on the eyes, increased body texturing quality, blemish removal, and more!
Image resolution downscale. Images will be 2K instead of 4K now.
New watermarks. Smaller watermark for both public and Patreon releases.
New poses and returning ones. Slowly bringing back all the group poses. Double fellatio, double cumshot, double everything!
So much customization. Time of day transition, scene transition, outfit copying, dialogue expansion, makeup presets, and more.
Going over the results of the satisfaction survey. Thank you to valliant gooners who didn't shy away from giving constructive criticism.
Themed sets and Transition poses delay. Working on getting it functional, but V4.6 took up most of my devtime this month.
Is Patreon in trouble? Is NovelAI in trouble? The payment processor mafia is tightening its grip. Payment Processor mafia. Exploring contingency and backup options.
V4.5 vs V4.6
Maintaining style while improving quality.
It is imperative to cheek poke Luci-chan during work.
Warning. Extremely long Monthly Report. ⚠️⚠️⚠️
I've split it up as best I can so you can find what's relevant to you, but there was simply too much to go over this time.
Time to ramble 🗣️🗣️🗣️
V4.5's success isn't stopping! I'm not letting go of the gas! Welcome to a massive number of new members~ 👀👋
The model is doing great! People are swarming in to the point where for the first time in 6 months I had to actually close the Premium Tier. Tons of members are throwing in ideas all the time - poses, potential narrative ideas, different scenes, thematic focus, and more.
The big surprise - V4.6!!!
Just a local ComfyUI upscaler?
Yes and no.
What does this change look like?
Here are 2 really simple examples. There won't always be this massive of a difference, but I' say the expected visual uplift on every single image will be around 25-50%.
This is one basic image from the Frieren set recently. Obviously a passable image with decent output. but we can easily notice tons of small issues
The eyes are failing to successfully detail the heart-shaped pupils. The makeup is rather grainy and rough. The eyelashes are rough. There's a lot of "scratch lines" around the nose.
The male penis is also a little bit rough looking too. There's just in general a little bit of texturing smearing happening which creates these sort of "smudges".
This effect happens on a lot of body textures in general. One of the complaints in sets was that they sometimes appeared a little bit "too greasy". While I enjoy sweat and more wetness in general, I agree there's definitely an argument to be made that sometimes it's a bit too much.
The idea is that this will remove a decent chunk of it.
So let's look at the improved version ⬇️⬇️⬇️
This is what I'd call a high level responder image. A lot of things were fixed, the makeup is much high resolution, the eyelashes better defined, the heart-shaped pupils functional, the penis texturing has been cleaned up as well. All in all, a massive success.
Now to highlight a little bit of the error correction in a way too.
Now we have an image from the Aru set.
Can you notice the problems with this image?
Tons of the same issues. 1. Shallow makeup 2. Thin lipstick 3. Texture smearing 4. Odd skin meshes 5. Eye color (slightly subjective)
And now the improved result. ⬇️⬇️⬇️
1. The makeup is a lot softer on the eyes, much more natural even if slightly less in quantity than before. Eyelashes are much softer and elegant looking. Eyeliner is very light, but equally soft and nice.
2. Lipstick has more volume and presence now
3. Texture smearing is mostly gone
4. The skin mesh while still present doesn't feel that out of place now
5. Eye color can be manipulated further, in Aru's case she has slightly multicolored eyes, but I chose to just demonstrate with yellow eyes to keep it simple. The iris looks much more stable, the pupil is far better textured.
Obviously the most important thing for me and it should be for any other AI creator is to maintain a style. Nothing gets people more apathetic to someones content if it just keeps changing over and over.
I don't want to become a generic shiny-skinned Illustrious trained carbon copy where every penis looks the exact same in every single frame with the saggy veiny balls and the thin long shaft with a large head. Or when every girl has the same proportions and even linen fabrics are shiny.
shots fired 💥🔫
Now... that doesn't mean that NO changes should happen, if quality leaves a bit to be desired then it's up to each creator to sit down and brainstorm ways to improve quality while sacrificing the least in terms of stylistic integrity.
F.A.Q.
How did you mitigate style changes from occurring?
While brainstorming how to avoid offloading more and more to a local diffuser, I came to the conclusion that the optimal solution actually is a mixed approach!
What I decided ultimately to do was to leverage the speed and flexibility of NovelAI, and then further refine the results with a local solution. Both working on an API level giving me as much customization as possible. I had to write my own custom nodes, which may not be super visible up there, but if you're a comfyUI user you'll probably notice the workflow is a bit odd looking.
Effectively this is an img2img solution where data is preserved and extracted alongside the base images and then used to help the local upscaler do its job with minimal errors. There's obviously a bit more to it than that, but we're getting into trade secrets at that point 😜
Currently the average failure rate is about 2-3%. Meaning that in the upscaling process if we have 500 viable images, 10 or so of those will probably be lost due to the complexity of the upscaling process. I'm working on mitigating this to 1% and below.
What does this actually mean for the visuals?
There's 2 major visual enhancements currently implemented -
The 1st pass is an overall visual standardizer in the form of an SD Upscaler at slightly higher denoise to make body texturing more consistent and clear up harsher drawn lines.
The 2nd pass is an eye area enhancing pass. The idea is obviously to further improve eye details, as well as provide volume and depth to makeup.
How does this affect the production cycle?
The production cycle will......................................remain unchanged!
That's right!
A big part of how I've set up my system is to allow for seamless and automated functioning. Speed and flexibility has always been #1.
There's currently been a consistent rate of 11 sets per week, with up to 44(!!!) sets per month for ONLY the Basic Tier.
Basic Plus+ members can see even more sets per month!
What other possible visual enhancements could you do?
Currently the main focus will be error-proofing the current upscaling model. The next solution I could potentially look into is making the backgrounds look better. But currently I'm almost reaching the point of diminishing returns in terms of upper-body visual clarity. There's still work to be done elsewhere, but at that point I also have to see what NovelAI might cook up.
KNOWN ISSUES -
Nipples sometimes are overdetailed even under clothing. Can be fixed with finetuning.
Random bright RGB pixels at times. Not super noticeable, but still finetuning it.
Double anuses. I clean these up myself, but if I missed any, oops.
Extra nipples. Again, I clean these up when I see them.
White pupils at times. Think of it as a bandaid solution for eye tracking
Penis precum texture meshing with the penis itself. It's struggling to separate the 2 things
Abs can be sometimes nonsenical on both the male and the female. Will be fixed in due time
Numbering of images in the Basic Plus+ tier can be slightly odd. They'll still work if they're viewed by name(ascending order) and/or by date modified. They can just start from a strange number.
These are part of the 2-3% error rate. Sometimes I can miss an issue with an image and it'll get in. I'll be working to filter out bad data like this over the coming days.
----------------------------------------------------------------------------------------------------------------
Image resolution downscale.
I'm downscaling from the current 4K image approach to 2K. Expect to see smaller filesizes (which some members, especially on mobile have complained about).
Will this have any visual quality difference?
For non-4K screens, no.
For members with 4K screens, barely.
This isn't a base level resolution change. It's only a post-processing image resolution change.
This was necessary to reduce filesizes and also to keep upscaling times within reasonable boundaries.
On the flip side, both promotional images on social media and Patreon stuff will now be equal resolution!
----------------------------------------------------------------------------------------------------------------
New watermarks.
Smaller and less annoying (the solo Patreon one is for Patreon sets, the other is for social media materials)!
Also a couple extra V4.6 images!
----------------------------------------------------------------------------------------------------------------
New poses and returning ones.
New monthly update, new poses!
Including, but not limited to - forward facing standing sex, pov neck grabbing during posing, group fellatio licking, group fellatio sucking, group fellatio cumshot, group sex posing, group sex posing upside-down, standin sex side pov against wall, suspended congress from below lifted sex, and more minor variation changes to existing poses (peace signs, diferent facial expressions, etc.).
----------------------------------------------------------------------------------------------------------------
So much customization.
Tons of new clothing, day/night cycles, new cafe/church scene, hilichurls, dialogue options, outfit copying functionality increased, and more!
Not much else to say here. The system can pretty seamlessly handle any sort of change in environment, lighting, and so forth. There's still more to do, but I'd say there's a decent bit of variety now within the set itself.
----------------------------------------------------------------------------------------------------------------
Going over the results of the satisfaction survey.
Disclaimer - To those who haven't filled out the survey:
Most questions range from 1-10 scale rating. 5 being a "doesn't really matter to me that much/I don't have a strong opinion on it either way" fallback choice.
Personally for me anything over an 8 rating is what I'd call a satisfactory result.
Overall satisfaction is great. Not much to focus on here, people are generally happy with some seeking a little bit of extra fulfillment. Let's explore!
Now bear in mind, this is in relation to V4.5, not V4.5a when I added a massive amount of new poses as well. I also added a lot of poses with this update as well.
In general people seem to be generally happy with the amount of poses but there's definitely a desire for more. This is something I recognized even before the survey poll which is why V4.5a tackled specifically that.
Currently people are craving more thematically oriented sets. I understand that, but it's not that easy for a creatively bankrupt person like myself 😭
This is a situation where it's not mechanical codewriting or the pose structuring itself impeding me from expanding on it, but rather more of a narrative writing perspective on things. I'm not a great storyteller, and people are asking for exactly that! I'll try my best, but it's something that'll be a bit slow to flesh out.
The split shows definitely a heightened interest in more sex angles and poses. Going into August I'll also include a few more.
Groping will likely see updates last.
You might have to download and zoom into this one, but lets go about it one by one.
Female detail - people are generally satisfied. I think V4.6 will further please people in that regard.
Male detail - in general a tad bit lower. Now bear in mind this is specifically texturing and quality of the visual, not size or anything like that. V4.6 will increase the quality of the entire male body significantly.
Makeup - Happy but not overwhelmingly so. Definitely one of the major reasons for V4.6 as I'm a huge makeup fan myself.
Body Proportions - People are a little bit on the fence here. A decent bit of 7's as well. I think in general this is a situation where I can't please everyone, since some just straight up prefer bigger asses and thicker thighs on everyone. I enjoy almost all shapes, so I try to keep the variety of toned, slim and curvy women as best I can.
Outfit Variety - People are generally happy here as well. At the end of the day request sets do take up a decent chunk of the released materials, so if the requester has simple tastes, the viewers may be left underwhelmed at times. This has already been somewhat addressed by the reintroduction of weekly polls and increased workload.
Facial expressions - Again, mostly happy. I think the variety of facial expressions could still be improved a bit more. A good way to convey a sense of the passage of time in sex is exactly through a gradual change of expressions over time.
NTR elements - Seems like most people don't particularly care about the NTR aspects. Now this is a bit deceptive, I'd say there is "hard NTR" and "soft NTR".
Hard NTR - poses where the viewer is directly interacted with. Outstreched arms towards the viewer, middle finger at the viewer, more direct facial expression towards the viewer (either disgusted or annoyed), talking to the viewer, etc.
Soft NTR - poses where the girl covers her eyes away from the camera/viewer, eye contact with the viewer, smug looks at the viewer
It's a bit too nuanced to put it in a survey so I just went with a more overall description. People are either happy with the NTR to most degree, or just indifferent to it. Which honestly is fine by me tbh.
Background details - Similar to the results above, people are either indiferent or mostly satisfied. This is just a fault of NAI in general, V4.5 seems to be a little less adventurous with the backgrounds and at times the character feels like she's being lit up by floodlamps, disconnecting her from the area she's in. I'll look into potentially including a local solution to maybe enhance the backgrounds, but we'll see.
Lighting - People are generally satisfied. I'm constantly fiddling with it. I've finessed the daytime lighting, but this is a general complaint not just with NovelAI, but also other SDXL trained models that AI just doesn't like low brightness.
Cumshot poses - This was where I wasn't super sure about the overall quality of the pose group itself. But people are generally happy with a few outliers with specific requests.
Feedback -
Instead of copy-pasting 22 different feedback responses to keep this document reasonable I'll go over the generally requested features. As to avoid any personal data being included I won't be including the actual link to them, but will go over them in order (you'll probably notice which one is yours).
More internal cumshots - Will add.
More cumshots/cum focus in general - Will add.
Pubic hair on the male. Interesting, will require a little bit more work to add, but I'll explore it.
More dominant/seductive female expressions - A bit harder to implement, but I've already started building the groundworks for this.
Pregnant epilogue poses - Will add in NTR-heavy themed sets.
Option for bigger and plumper asses - This is a bit harder since I don't want to betray peoples expectations of a chracters proportions. Requested sets may have this feature, but Poll Winners most likely won't. Will think about some ways to incorpoate this more.
Different dialogue directions (e. cuckqueaning over cuckolding, etc.) - This is my weak side. Creative work isn't what I'm good at, but I'll try to expand the system to allow for this. It's not that the feature doesn't work right now, it's just hat I don't know what to make them say lol. If you have dialogue bits you want to see, let me know.
More active sex poses - This is an interesting one. The reason why I picked NTR to begin with wasn't that I'm a particularly massive fan of it or anything, it's just that with NTR you give yourself an excuse to frame the pose in such a way that most of the female is visible. Let's examine a from top PoV perspective of a doggystyle pose. What do you see? Not much really. The breasts are mostly out of vision, the pussy/pubic hair isn't visible as much, the facial expressions of the female are lost at times, etc. A lot of dynamic in the "camerawork" is lost due to that. I'm not saying I won't add any new Faceless Male PoV poses, it's just that they'll pretty quickly hit the limit since right now an NTR approach gives me tons of different angles - from side, from below, from above, from front, from behind, etc.
More heart-shaped pupils - Added a few with this one, will add more.
Penis can be a bit large at times - There's a lot of reasons for why this happens, sometimes the penis increases size to match a larger framed females size. Sometimes the girl is small so the guy gets to take up more screenspace. Sometimes it's bonus content the requester wants. I in general prefer a bit larger than "large" size, but I can agree there's moments where we're talking like dislocated jaws and internal hemorrhages xd.
Penis worship/fellatio more penis kissing - This was an oversight on my part. I'll work to add those very soon, lipstick marks as well!
Sex scenes more smiling at camera - This is a good point to make. I think transition poses here would allow for a little bit of "downtime" in between the crazy sex. I agree the poses are a little bit nonstop there and some change of facial expressions will be good.
Set cohesion - sets having a proper beginning, action, and end. I agree with this, definitely conveying a sense of time progression is very important and the introduction of transition poses is where this will help immensely since they'll also often act as "reset points" for the events. For example the girl just went through a hectic facefucking pose, her hair and makeup is ruined, she needs some time to "clean up" and resume the activities. That's definitely something that will happen in August.
Smaller male models and male hairstyles - This is a bit funny but yeah, the males at time appear a little bit too faceless and bald. The body size is something I probably won't change much, since a lot of head positioning data is tied to there being a substantial height difference. But the hairstyles is something I can include relatively easily. Not quite sure how to procede with this one just yet, but thank you for getting me to think about it.
More NTR elements - Wedding rings, mocking expressions, etc. Will add with themed sets.
Public sets lacking variety - Weekly polls will explore fun new outfit options. But in general I imagine most people love their anime waifu and her signature clothing most of the time. Kind of unavoidable. Some outfits are so good I don't want to change them too.
Kissing poses - Will be included in themed sets.
More annoyed/glaring expressions - Will be included in themed sets.
More anal sex representation - This is a hard one, it's not that I can't it's just that it would make the image count explode. I'll explore some solutions maybe for non-requested sets hey'll have a mandatory 80/20 split at least to allow for anal sex as well. I'll probably do a poll on this later on.
Unsurprising to most. The most valuable feature is the most narrative directed one. The people are craving a story and I'm failing to deliver on it still.
At least I managed to start bringing back group content.
That's all for the survey results. Thank you to all the participants and congratulations to the winners!
----------------------------------------------------------------------------------------------------------------
Transition poses and themed sets delay.
Not much I can say here other than just apologize. I'll go over why this feature was delayed a little bit after this.
Transition poses will come first with themed sets a bit after that.
V4.6 simply ate up my devtime and I'm not a super creative person either, so sitting down and brainstorming how to fit these into the rotation has been a bit harder than expected.
The reason for V4.6 has a fun side to it, but also a slightly darker undertone to it which you can read in the next part.
Another issue is unfortunately there's simply too much work. I've began downsizing the Premium Tier to have more free time. With time things will calm down a bit. The most I can do now is just plead for your patience 🙇♂️
When should we expect the first batch of transition poses at least?
August for sure. I have some ready, but they're not finetuned enough just yet. I can't give a strict date, but it should be around maybe the 2nd or 3rd week.
----------------------------------------------------------------------------------------------------------------
Is Patreon in trouble? Is NovelAI in trouble? 😟
This is an unfun conversation to have, but it's one I have to address. I mentioned above a few times that V4.6 wasn't brought upon only by the pure desire for higher visual fidelity (even though that's the main point of it).
In reality it's also a backup/contingency plan ⚠️IF⚠️ NovelAI runs into problems with Payment Processors like Visa/Mastercard.
Why would that happen? What's going on?
I advise checking in with you favorite gaming news Youtuber as they've probably also covered this, but recently there's been a massive crackdown on NSFW content by Visa and Mastercard. Instead of me going over it here, check one of those videos first and then come back to hear my side of things.
These changes can potentially affect either of the 2 platforms -
Patreon - they could decide to just ban NSFW content altogether like itch.io did. Or severely limit the capacity in one way or another. Luckily the solution here isn't as awful, everyone will just move to another platform (either subscribestar, or fanbox, or something else). While that doesn't solve the issue permanently, it'll be a temporary fix.
NovelAI - this ones much worse for me. It'll basically kill my style. Obviously all my prompt data and system logic is still intact, I'll be able to offload the production process to a local solution and become a part of the fully local community. This can affect both the speed of sets as well as severely change the visual style. I'll obviously try to replicate it as closely as possible, but training LoRa's is a brutally time consuming process, so initially a dummy style might be used. Rather not think about it too much ;_;
What can people do?
Again, check with your favorite Youtuber and you'll see there's various initatives people are undertaking to somehow curb this ridiculous spread of online censorship.
This is the quickest summary I can give you about how this will relate to your enjoyment of my content and others' on the Patreon platform. While currently there hasn't been a formal statement by either platform on how they're going to approach that discussion. It's highly unlikely either would decide to fight a battle against Visa/MC lol.
This is the extent of what I'm going to comment about this subject on this platform. The topic is interesting, but very extensive, exhausting and can turn your mood sour rather quickly. I want to keep the content here free from the plights of the outside world and focus on erotic NTR content!
If you want to chat up more about it, drop by the discord at https://discord.gg/3PF4M6Ug5J. (JOIN NOW JOIN NOW JOIN NOW JOIN NOW JOIN NOW JOIN NOW JOIN NOW JOIN)
----------------------------------------------------------------------------------------------------------------
Tech babble for a few of my members who care about it.
I hope V4.6 will be received just as well as V4.5 was. I've always had a chip on my shoulder that my work wasn't as clean and crisp as other people in this sphere. Obviously NAI will always be visually weaker than a local solution, since the difference between generative times is probably 2s vs 20s. But I still wanted my stuff to look clean and shiny.
I think finally it's gotten to a place where I can just say it looks satisfactory. I just hope with all my heart NAI V5 will finally increase the base resolution by at least 20%.
This will sound narcissistic and arrogant probably on top of being wrong most likely - but I'm really happy I've managed to build a fully interconnected system between NAI's API and ComfyUI's API.
A big part of this project is software development at a product planning level. Something that isn't my specialization. My tech experience isn't in that sector. While my background did help, this was definitely fun, albeit time-consuming to build.
There's a few tech nerds here who asked how it works so I'll go into it a tad bit now -
The ComfyUI app sucks, don't use it. Use the API, write a few custom nodes for handling the interconnected data like metadata extraction and sanitization. Leverage the speed by going baseline lowres on NAI's side, then diffuse and upscale on local with medium high or low denoise (based on your style), samplers don't matter as much since you're doing img2img, but still fiddle with them a bit. Similarly tile cleanup and seam fixing doesn't matter as much. Use a regional detailer, leverage its CLIP encoder to handle more nuanced cases, don't go high steps on either, you don't need to.
Handle all your character building logic locally nesting it into the NAI API.
Don't overcustomize, trust in the AI at times.
Buy more RAM. Yeah, 32GB isn't enough.
Save every metadata file, don't think you won't need them again.
Make sure every prompt is perfectly written. It matters for the post processing upscale.
I can't believe a lot of creators are using this approach which is why at times I check up on the competition and outside of the literal top dogs with thousands of members, not many creators are capable of putting out 44 per month fully customized 450+ image sets.
----------------------------------------------------------------------------------------------------------------
Crazy milestone! I want to thank everyone one more time!
😚😚😚💖💖💖
Whew. Finally done huh?
Man that took a few hours across several days to finish writing. If you actually read all of this schizo blogging by some random gooner on the Internet you're a beast tbh, thanks.
Comments
IAMDAROYAL
Seems like a good update overall c: also pretty dope to see you want more feedback from fans
Caspiel Rex
I think for the areas where you want to give people the most choice in their requests, it's better to have that obvious in the request form. I know you will do what I want as best you can when I write something in the box, but other people might not, so asking for "can they have narrower hips" or "can the penis be this size" isn't as apparent (I know for me the body shape is pretty important because I like to see the characters as original as they are). Maybe give example images of body types and let them check the box or to just use the default character proportions. Ofc there is the other end of decision paralysis, but I think the fact that people are paying money is a good motivator to figure out what you want. If you want more specific details or feedback I'm happy to provide it in DMs.
