Mom's Texts

Horny TL;DR: let's open up your phone and go through some of your messages, let's see...

You can find the screenshots in the release archive here (all 4 images are also attached to this post):

Releases\2025\T-Texts

Hey everyone, was playing around with one of those fake screenshot making sites and decided I liked what I made and thought I'd share it. You wouldn't know it, but I used to write erotica (on Literotica) back in the day. I'll never admit to being great at it, but it can be fun. I also want to write some lore and backstories about my "characters" sometime. And give them real names probably. A few people at least have requested more lore and backstory, and I am happy to oblige to some extent. I say "characters" since their appearances morph and change so much due to the inherent nature of AI. Sure, I could start training LoRAs on them or do face swapping/InPainting to decide on one appearance, but 1) that's a lot of work and 2) I like the variety and "evolution" that the characters take on and 3) the variety doesn't lessen the level of "hotness" i think the character has.

I finally got FramePack working at least. I can't say I have it really dialed in yet, still needs lots of setting adjustments. But, at least I'm not running out of memory during generation now. FramePack is supposedly leaner than Wan 2.1, but I'm not seeing that yet, could also be that I'm using a very well optimized workflow and GGUF quantized model for Wan (FramePack is very new and the lowest available quantization of the model is FP8). But, the neatest thing is that it generates at higher resolution and longer lengths with only marginal VRAM cost increases. I was doing 10 second videos in 35 minutes yesterday. I'm going to try some 20 second ones today. It is also neat that it handles physics and appearances different than Wan 2.1. Dare I say more realistic at times since Wan seems to drift quickly causing my images to look more computer generated and 3D (or at least much of the detail of the original source image is lost almost immediately). But, the physics and motion are a little weaker. And I haven't tried LoRAs yet and the prompt following seems poor at my current settings. So, early days. I'll try to post some of the experiments soon, some of them are quite high quality.