[Amethyst Glow] Size focused cuckolding, mostly interracial

[fantafan]

Pretty decent cuckold stories, all focused on bigger cocks than hubby, mostly/all interracial.

 

[Nearby]

Thanks for the books. Checking on the author's page, I can't help but feel very sus that these books are AI-generated considered they've put out more than 10 full length novels in less than 2 months lol.

 

[fantafan]

I suspect they're all reissues. Seems to be happening a lot given the ease of generating new AI covers. But what do I know? They don't feel AI generated at least the few I've read.

 

[SmeeJ]

There's a few weird bits between some chapters that suggests to me they come from a Patreon style chapter by chapter issue. e.g. discussing what's coming up in the next chapter. eg. this passage in "Cuckolded On The Island":

Okay, I'll move into Chapter 11, focusing on these "accidental glimpses" and Jason's unsettled reactions. Given the 10,000-word target, I'll aim to draw out these scenes, focusing on Jason's internal monologue, the subtle (and not-so-subtle) body language, and the building tension.

**(This will be another substantial section. I'll write as much as I can in one go, focusing on the first "glimpse" – Beth looking at Marcus shirtless – and then moving towards the second incident if space allows, or setting it up.)**

Suggests to me a chapter-by-chapter release that wasn't quite cleaned up.

 

[O_P_T]

Weird, the covers on Amazon & Goodreads are in color, not B&W

Edit: Never mind, it makes sense since fantafan's using a B&W kindle to get the books.

However, it is strange that some books the author is "Amethyst Glow" and for others it's "Glow Amethyst"

 

[fantafan]

When they misspell Honeymoon on the cover page...

 

[yano2mch]

Thanks for the books. Checking on the author's page, I can't help but feel very sus that these books are AI-generated considered they've put out more than 10 full length novels in less than 2 months lol.

Hmmm... Been playing with LLMs and SillyTavern. It doesn't take long to generate a lot in a directed or direct roleplay method. But depending on model and story and how well it was done, it can feel low quality, or need rewrites in sections.

They don't feel AI generated at least the few I've read.

Certain word patterns and phrases are FAR more likely to come up in AI written... Starting to feel cliche when i see 'don't think for just one minute' repeated for the 8th time...

The 70B models i've been playing with feel eerily close to what a good RPer would have replied with. And with a good workflow, character and environment notes set up and proper direction...

Who knows. If the quality is there that's what's important.

 

[KidneyCracker]

When they misspell Honeymoon on the cover page...

I cringe when I see stuff like that.

 

[fantafan]

Playing with initially free chatbots I haven't been that impressed, but I'm sure that's not state of the art.

 

[yano2mch]

Playing with initially free chatbots I haven't been that impressed, but I'm sure that's not state of the art.

I doubt i'll ever touch 'state of the art'. But i am tinkering with larger and larger models that are available locally.

Some 70B models output pretty darn well too, good enough it doesn't seem like a computer wrote the replies (at least in a RP chat); I'm trying out a Q2 235B model that takes a lot of RAM and it's output is very promising..

 

[O_P_T]

I doubt i'll ever touch 'state of the art'. But i am tinkering with larger and larger models that are available locally.

Some 70B models output pretty darn well too, good enough it doesn't seem like a computer wrote the replies (at least in a RP chat); I'm trying out a Q2 235B model that takes a lot of RAM and it's output is very promising..

So you're saying it's getting close to be able to pass a Turing Test? At least in a limited sphere of questioning?

 

[yano2mch]

So you're saying it's getting close to be able to pass a Turing Test? At least in a limited sphere of questioning?

Oh i think the 8B and 16B models can pass the turing tests; Truthfully it's answers seem quite good compared to a number of humans. Actually I'd say a lot of people are too focused on their political ideology (educated into imbecility), to which i'd rather have an intelligent bot than them.

But the larger models have more imagination and keep track of details better. I'd been chat RPing since mid 90's when chatrooms became a thing, so yeah the larger models are becoming quite nice.

 

[7hR337R0723]

I doubt i'll ever touch 'state of the art'. But i am tinkering with larger and larger models that are available locally.

Some 70B models output pretty darn well too, good enough it doesn't seem like a computer wrote the replies (at least in a RP chat); I'm trying out a Q2 235B model that takes a lot of RAM and it's output is very promising..

How are you running 70B and 235B locally? Doesn't it require a system with multiple H100s or some other powerful cards?

 

[yano2mch]

How are you running 70B and 235B locally? Doesn't it require a system with multiple H100s or some other powerful cards?

This is more a RAM (and GPU VRam) problem than it's anything else. All the models i use are quantized (scaled down in precision). Q8 is the highest the GGUF files. That means most of the models i use take up about 40Gb in filesize and in Ram. Though the really large models, even a Q2 in the 235B model eats up like 70Gb ram.

 

[7hR337R0723]

This is more a RAM (and GPU VRam) problem than it's anything else. All the models i use are quantized (scaled down in precision). Q8 is the highest the GGUF files. That means most of the models i use take up about 40Gb in filesize and in Ram. Though the really large models, even a Q2 in the 235B model eats up like 70Gb ram.

Ah so you're offloading some layers to main RAM. Makes sense. Doesn't that make it very slow though?

 

[yano2mch]

Ah so you're offloading some layers to main RAM. Makes sense. Doesn't that make it very slow though?

Yes. The GPU is much faster. But the GPU only has so much RAM to work with. The programs are made to run in both, so you offload everything you can to the GPU. If you're lucky (or rich?), the whole thing.

Honestly getting 2-3 tokens a second is a decent speed, and higher is better. But if we want to talk a lot more, probably move to the Text Generation thread i made where i cover models and experiments in this topic. We've gone quite off topic for too long here.

 

[RodinParis]

Pretty decent cuckold stories, all focused on bigger cocks than hubby, mostly/all interracial.

View attachment 1659554 View attachment 1659555 View attachment 1659556 View attachment 1659557 View attachment 1659558 View attachment 1659559 View attachment 1659560 View attachment 1659561 View attachment 1659562 View attachment 1659563 View attachment 1659564

My Wife’s Raging Hormones was totally different to the others so far and really intense cuckolding

 

[tstplt]

Weird, the covers on Amazon & Goodreads are in color, not B&W

Edit: Never mind, it makes sense since fantafan's using a B&W kindle to get the books.

However, it is strange that some books the author is "Amethyst Glow" and for others it's "Glow Amethyst"

You can find these books on Smashworkds, the covers are also in color.

 

[tstplt]

Thanks for the books. Checking on the author's page, I can't help but feel very sus that these books are AI-generated considered they've put out more than 10 full length novels in less than 2 months lol.

I agree. I suspected AI generated too, but overall the stories are pretty good. But I have also not been able to find any information at all on the "author" Glow Amethyst. And the versions I have need intense editing. One chapter was duplicated, events are referred to that did not actually occur earlier, etc. But the style of these books is unique, quite a lot of buildup, from what I read good and plausible character development (given the genre.) Oh well, my thoughts.

 

[fantafan]

A few of the new ones. Cost money so I didn't buy everything. Feel free to contribute.