DuxVector

V4 incoming, New $10/month Tier, Subsystem announcement, and Price Update for Request Tier ($15->$35 per request. Starting after the next billing cycle. 1 month grace period!)

Added 2025-04-30 20:35:00 +0000 UTC

Check out the cute Luci-chan comic below~

TL;DR (SHORT VERSION) of upcoming changes -
V3 SUPPORT AND UPDATES WILL CONTINUE. I will maintain both models. They will NOT mirror functionality 1:1 though. Some features will be model exclusive.
NEW $10/month tier (Basic Plus). Will have access to ALL public works (including promotional materials), as well as all unreleased public sets with NO WATERMARKS. More details below. Will be up in the evening of May 1st. UPDATE: 05/01/2025 - Will host on Patreon over MEGA.
Request pricing increases to $35 per request. This was unavoidable due to the massive increase in necessary production time and maintenance of current code.
IMPORTANT - All Request Tier members will receive a 1 month grace period where they are still billed the old price and will still receive 2x sets as per the terms when they were billed should they wish to continue their membership.
Don't worry, you won't get billed the new price out of the blue and only get 1x set to request.
Standard Tier and Premium Tier will be consolidated into one. "Extra" requests will be included as an on-demand payment option for those who wish. Will not always be available based on schedule.
Due to the significant price increase anyone who feels financially overwhelmed by it, please, do not feel ANY hesitation in cancelling your membership. If anyone has been billed unexpectedly in the future, please, reach out to me and I will issue a refund on the very same day. You put your trust in me, now it's time for me to return that trust. I enjoyed working on your vision and hope to see you again soon!
As a further bonus this also extends to V4 requests too! Think of it as a super discount for being at the cutting edge of technogoonery! I advise waiting to see how V4 looks like first in the following week before taking advantage of that though!
Words cannot begin to describe how humbled I've been to receive all this generosity and trust from dozens of people across the world. From the bottom of my heart to those who decided to gamble on my skills and found them satisfactory, I extend a most heartfelt and sincere "Thank YOU!".
V4 entering large scale production from tomorrow (May 1st). It's visually not as stunning as V3, but can handle a lot more data reliably, and most importantly newer characters (Up to January/February 2025-ish). More details below.
V4.5 is in active development by NAI. They are aware that V4 is underperforming visually and seek to improve the visual quality. More details below.

OK. That wasn't a "SUPER" TL;DR version this time, just a TL;DR.

Off to rambling!

Whew... Where to start...
V4 is here.

I had to overhaul quite a few systems and rewrite a lot of poses to fit the new POST logic requirements when I work with the API. It's still rough around the edges and only works with Bedroom scenes right now. But scaling up that part won't be difficult. Expect to have at least 2 indoor and 2 outdoor scenes ready by the end of the first week.

The visual side of V4 leaves a lot to be desired, but in the next coming days I'll properly showcase a bit of what it's capable. Pretty much everyone is dissatisfied with it in terms of the visual quality, and the quick push for V4.5 on NovelAI's side means that they're aware of that. I'll continue to optimize and improve the visual side of things of course. The end goal being a near identical replica of V3.

V4 Database cutoff

It's a bit hard to estimate this time around since the API "suggest" metrics aren't using Danbooru numbers. While it's still using Danbooru as a source, it's no longer the exclusive source (which is why the quality isn't up to par).

https://image.novelai.net/docs/index.html#/default/get_ai_generate_image_suggest_tags

I apologize for the messiness. The NAIV3 Tag search tool is as the name implies, only functional with V3. So we'll have to use a roundabout way to check for V4 characters and tags.

https://image.novelai.net/ai/generate-image/suggest-tags?model=nai-diffusion-4-full&prompt=

---------------^-------- You can include anything you want after the = to check for you favorite new character.

https://image.novelai.net/ai/generate-image/suggest-tags?model=nai-diffusion-4-full&prompt=evelyn

The above character is Evelyn from Zenless Zone Zero.

Her first piece of art on Danbooru was on December 13th 2024. And she still has a prompt power of 5.6k which is a lot. We can somewhat safely assume the cutoff is sometime around Late January/Early February 2025.

Price increase reasoning

For those who may have missed it, it's detailed in the post below.

https://www.patreon.com/posts/125698925

This month alone I handled 33+ requests. Each request can average from 2-4 hours to flesh out from beginning to end, including the time it takes to set up the correct prompt details, edit, focus inpaint, upscale, host, communicate, and deliver.
I uhh.... Haven't had a day off in a while....

This is excluding any time I spent on refining V4, cleaning up junk code, integrating new features, adding new poses, testing new poses, etc.

While I'm enjoying the continued increase in scale of this hobby project, I have to also make it reasonable.

V3 and onwards.

V3 will continue to receive updates. I've just recently included new focused group poses, as well as a bonus to the poses with added hair pulling. Constant micro adjustments have been made to many of the more brighter scenes such as the School and Beach ones, making them a lot nicer on the eyes.

I'm working on adding a bunch of selectable toggles when people make their requests such as less make-up, more make-up, less toned body, more chubby body, etc. Will probably be ready sometimes in the middle of May.

$10/month Tier information (Basic Plus)

Basically a strong sidegrade to the $5/month tier. Allowing for people to have instant access to all public sets with no watermarks. As well as exclusive access to all previous and future promotional materials with no watermark as well.

Of course, the Premium Tier (Request Tier) will naturally have access to this as well.

The idea behind this tier is to set free the sets stuck in perpetual limbo. Some at this point have gotten either so old that they're missing core features (focused inpainting), or are missing a lot of poses, or are just in general niche characters which couldn't snatch the spotlight away from everyone's favorite Azur Lane girl of the day.

UPDATE 05/01/2025 - Will simply host them on Patreon.

V4.5 News

Not much to say here. Just the average copium that it's gonna get better. I've been browsing various social platforms to gauge public opinion on V4. And even 2 months later people are still upset with it. That's probably what must've prompted the quick panic at the NAI offices.

Allegedly the Curated version will be available for public testing within a month or two. And people are expecting the Full version to go live much quicker since it's a smaller update as well. Hopefully Late June? Cope...

Other fine control include the ability to pick lipstick color as well. Potentially symbol-shaped pupils as well (hearts and such), more specific makeup (specific eyeliners and eyeshadows), tattoos, and more.

Something V3 couldn't do at all...

Off to Luci-chan comics~

HELLOOOOOOOOOOOO LUCI-CHAAAAAAAAAAAAAAAAAAAAAAAA----
Oh. She's on the phone with the AI Overlord (Praise the Omnissiah, the hallowed Machine God).

Hey, hey Luci-chan~

Didn't you hear? You're going to get digitally dismembered, torn apart limb from limb, and dissected meticulously by specialized AI chips so that you can work in multiple models now!