[Access Tier Release] Shinobu of Demon Slayer | Variety Sex [Lightning LoRa Gen 2] (Patreon)

SOOO Tomorrow Access members are in for a nice treat!

4 whole Variety Sex sets are being released! One is pretty much Shinobu again, but not with the lightning model/LoRa, so next time the quality will look better! :D

All 4 sets are about 90+ images each!!

======= Little Announcement and some Technical info ======= 

Also, I can confirm I am finally getting a proper fully upgradeable Desktop!! For now, it is estimated I will get a %30 increase in speed, it is a 4070 TI Super with double the Vram from the 3070 8GB so 16GB! This should help with retaining good performance when using larger Lora and SDXL models while doing hires fix. Larger models mean more accuracy, ( Discovery made on Training on Characters.) this also applies to SDXL models too.

Generally, if you use a lot of LoRa models or you have large ones (or larger SDXL Models) it slows down generation (depending on the Vram size), thankfully A1111 Forge has been incredible with performance so it's helped A LOT over the original A1111 Web UI.

My goal after this is to hopefully save up for a 5090, as the info has been leaked by a credible source, atm it would be a good little while though.

The 5090 is estimated to have a 412-bit bus width(the GPU I have now and the new one with my desktop is 256-bit), it has 28 GB of VRAM, and 50 watts more than the 4090,

The only thing holding me off from going to it is the issues that the 4090 had on the power connector, which was getting burnt so I want to at least wait and see how it goes.

One thing I do like the most about the 50 series is the Blackwell architecture and hopefully even more geared to consumer AI generation.

(40 series came out in the very early days of Stable Diffusion, where SD was not even widely used at all, up until November-December 2022 when SD 1.5 Came out). SD was released in July of that year, so 2 years ago!

So my hope is they developed the new series with consumer AI in mind, considering Nvidia even made a GPU fix to an update that messed with the performance of SD generation back in August I believe, and they took notice and made a fix for the update. So Nvidia is aware of the AI art community and of course the Local LLMs and many other AI model communities that sprung up.