Solo RP: Thousand Year Old Vampire – Angel Fall – Bonus Scenes #1 – The Yamori Gecko

Content warning: This post contains two women expressing love for each other. (Seriously, does this really need a content warning?) Nothing explicit: Just the insinuation of touch and a tumble in bed at the end. ChatGPT wrote it, and I did not push on any guard rails, so it has to be tame, right?

It also comprises of inadvertently long walls-of-text commentating on the whole TYOV narrative arc and a rambling philosophical reflection on AI, writing, creative experiments and literary digressions on “found language” and “plagiarism” before actually getting to the lovey-dovey bits.

Which might be the true content warning.


By the end of our Thousand Year Old Vampire playthrough, it became obvious that the long term narrative arc across the prompts was ultimately a love story.

Of the soulmates lost in time trying to find each other variety.

(Except in this case, one of them stopped trying. Decided to go all Gangrel and walk in the wilderness, and seek Golconda in Daoism, to steal some Vampire: The Masquerade references.)

The potential was already seeded at the start with the reason for Astral’s fall.

What I left up to fate and the dice rolls was when Iguana might potentially step in, whether either would hurt the other to the point of an unsalvageable relationship (by following any TYOV Prompt directions) or if there would be some kind of buddy cop partnership early on or in the middle of the story.

The story the dice (and my subconscious) decided to tell was an extended journey of walking alone, distantly in contact with examples of different kinds of love and affection.

  • Rami was the original mortal encapsulation of what Astral and Iguana’s relationship was and could have been, except wrapped around the inevitability of early loss once Rami aged (and our vampire didn’t even get to that possibility, once kicked out with a flaming sword)
  • Flame was the complicated kind of sibling love
  • Mehrun was the uncomplicated kind of love of and for children
  • Ishraq was obsessive love (and if we really stretch the metaphor, Khalida may have been…narcissistic love?)
  • Salima was one-sided love and love generous enough to let go
  • Chimeko and Tenzin were… perhaps a representation of witnessing couples’ love from a outsider third-party single perspective?

And of course, when Astral gave up being bitter about banishment and thinking herself unworthy of staying in one place for long, stopped striving so hard and accepted (imperfectly) balancing on edge of contradictions, it was time to re-introduce the possibility of romantic soul-deep love.

Just in time to catch her at the point of her most vital dilemma, when the old ways of doing things no longer worked for her.

(Unrealistic? Sure. But that’s fun of fictional storytelling. You can make up narratives more satisfying than real life.)

Speaking of satisfying… it’s terribly unsatisfying when a love story ends without sufficient time lingering in the epilogue and the aftermath.

-I- wanted more, anyway.

(And if solo roleplaying can’t do wish fulfillment, we are definitely investing effort in the wrong pastime.)


Here, we’re going to experiment with an interesting spin on “creative writing” in conjunction with ChatGPT.

That strange supposedly “know it when I see it” category of words that some people term “AI slop” is generally a human cut-and-pasting generated text directly from an LLM (and oftentimes passing it off as their own.)

I actually agree with that latter bit, that such things should be transparently declared, as in the extent of AI use and how it was used.

But a) do these people really know it when they see it?

Author Mark Lawrence recently ran a quick flash fiction experiment where he surveyed the general public (well, people who read his blog or frequented certain subreddits during the duration of the vote, anyway) and asked them to see if they could tell apart 4 AI-written short texts from 4 human-written ones.

The results were… telling. Not that many people could.

And on an enjoyment scale, for that particular test, AI produced texts on a higher enjoyment level than the humans did.

(Very possibly due to ease of readability and/or more common general appeal tropes – AI is trained by majority standards, after all – while the human authors appeared to be leaning more experimental / unique, which may have seemed like hallucinations to some readers.)

I got 5 out of 8 right. So I’m no better.

b) are there instances where AI-generated text can be helpful though? Assisting people without the ability initially? Or as scaffolding for learning, perhaps?

Once upon a time, I remember some writing advice that went along the lines of advising new writers to pull open a book of their favorite authors and start copying their lines. A form of slow study of sentence and scene structure, and how these writers were expressing themselves.

Imitating their voice and blending them into a mix and maybe eventually finding one’s own voice while doing so. Imitation being the best form of flattery and all that.

(Obviously, not plagiarize and attempt to pass it off as one’s own work, which would be a violation of integrity, let alone copyright.)

Could learning from the many different ways AI is able to say the same thing in readable and grammatical ways be a form of the above?

Might new writers being able to exercise critical thought and subjective judgement in identifying the good bits in AI-generated writing (as opposed to the terrible bits) be a method of training their own literary taste and voice?

c) are there similar things in related families that are considered “legit?”

The first thing that came to mind is a cento in poetry. It’s a poem made entirely out of verses taken from other poets.

(I wrote a cento out of Edgar Allan Poe verses, way back in 2014. To reflect on the MMO burnout cycle and the reasons why we play MMOs.)

Following that thread brought me to concepts of bricolage and citational fiction and the literary supercut. Except, as the previous linked article says, there are no common terms for it. “Found language,” “cut-ups“, commonplace books and collage writings all mix in this genre, and mean different things to different people at different times.

(I’d also volunteer the concept of “mashups” in music and song remixes. I listen to nightcore and Youtube covers all the time – some of them are significantly elevated over the original.)

People can steal words from people and it’s either plagiarism, violation of copyright, or really high falutin’ experimental literary writings apparently.

Steal words from an LLM that was doing the same thing with originally people’s words and it is plagiarism, violation of copyright AND slop? 🙃

(Let’s be clear: Common definitions of plagiarism, as defined by scanning the first page of Google results, seems to be more along the lines of presenting the work as your own, especially without attribution. That I agree with, more or less, and think transparency and proper attribution is the way around that.)

and d) Can it be done well? How much time and effort does it really take?

As an experiment (with full transparency and proper attribution) and personal writing practice exercise, I thought I’d give it a go.

I’d already generated a number of versions of the same scene in ChatGPT by using the edit button and letting it re-send the same prompt.

That prompt is the human baseline, providing scene direction and steering the LLM along a common path.

It then returns a potential prose version of the scene, remixing language in different ways (and in the case of GPT-4o, tossing in some ‘creative’ aka ‘hallucinatory’ additions that are great for people seeking novel ideas and infuriating others who’d prefer GPT follow instructions to the letter. Try GPT-5, maybe, I’d say to the latter, but apparently it’s still kinda dumb. Oh well.)

For personal entertainment, I just read all four GPT-generated versions and enjoyed the fun bits, and discounted the weird out-of-context bits.

For this “writing” experiment though, I’m going to steal all the fun bits and blend them together in a new mashup. Is the result any better than the sum of its parts?

Highlighted in yellow above is all the bits that I liked, from each GPT-generated text.

We’re going to re-arrange them in a logical sequence. Clean up the tenses to one unified version. Make the smallest tweaks we can to stuff that really sounds wrong. (Sakura leaves, for example. A lot less cool than sakura blossoms or petals.)

It’s not really creative writing, if writing is solely defined as the generation of words on a page. I struggle with the concept of claiming the above as entirely “my” writing. 90% of it is not.

BUT it is an exercise in “creative editing” or “revising.”

Which… is also part of the craft of writing, is it not?

It took easily two hours or more, sitting there reading each GPT text, carefully highlighting each sentence I liked, holding them simultaneously in my mind, figuring out the best of four versions to offer up in a sequence that flowed and made sense and read well.

Then I also had to cut-and-paste each fragment or re-type it all, and ensure the sentences flowed well and didn’t have weird tense switches.

I am still as proud of it as I would be a cento. Effort went into it. Human taste.

It’s 100% my choice on which of the sentences I used in the end. Based on the yellow highlights, it looks like an even 30-40% from each GPT text, surprisingly.

The words I used to glue the bits together also did come from my brain. It’s 100% my concept and early stage direction for the scene too, as can be seen from the initial prompt.

It’s the stuff in the middle that is 90% AI-generated sentences.

(Which, according to that Mark Lawrence informal survey, is a LOT more readable than some human-generated sentences. I can buy that. Run-on sentences and verbosity are a characteristic of this blog, and I’m getting rid of that only over my cold, dead body.)

If someone wanted to learn to be more readable though… is it that wrong to pick that up from AI?


Same idea with the second scene. Six GPT texts this time. A few varying prompts.

The first was produced by GPT as a spontaneous continuation of the first prompt. The next two was a short, sparse prompt just asking GPT to riff on the “making up for lost time” comment by Iguana.

The final three were from three cycles of a directed prompt about Astral’s Marks.

And now we mash. Mix ’em all up into one.

Cento?

Or slop?

You decide.

I’m already happy. Ship achieved.

Next up, some narcissistic commentary with my co-GM on the whole TYOV experience and realizing that I can do “additional scenes” with myself, rather than “deleted scenes…”

Disclaimer: A Photoshop tweaked AI-generated image was used as the feature image for this post.

Solo RP: Thousand Year Old Vampire – Angel Fall – Part 26 & End

Previously, our vampire bogged down in severe bureaucratic red tape (plus prejudicial bias) and tried to invoke a miracle to make everyone feel better… It didn’t work. Belief stopped being the fiat currency it used to be.

Today we roll an 8 on a 1d10 and a 5 on a 1d6.

Three steps forward into a game-ending prompt:

Wow.

What does this mean?

ChatGPT-4o has been a partner throughout this solo RP experience and the last couple of responses from Part 23 onward have been, on the whole, surprisingly near to target.

I assume it was building off the context of the contents of the entire chat window, since we were playing this game / storybuilding exercise in one contained space. By now, we’ve accumulated quite a lot of relevant text in that space.

So I hand off (partially) to GPT to propose some ideas for what this ending might look like.

Some version of Option 2 was definitely in my mind as a given. Why would I have looked so hard for an opportunity to sneak Iguana back into Astral’s orbit otherwise?

But the proposed Option 1 really surprised me. I’d long forgotten (just like Astral) that Library Scholar was one of her original Skills, especially after TYOV prompts told us to throw it away.

It blended that callback, her introvert personality and mixed in the tiniest echo of canon Astral – a hacker in our cyberpunk-flavored supernatural hero fantasy Tavernpunk Tales of Ares setting.

And then came up with something utterly novel, a secret archivist, a librarian of a hidden collection.

Thanks, GPT, for taking my prompt so literally that you fucked up the POV perspective.

(I suppose I should have left out the quotes, or said something like “Include the idea that she remains…”)

It’s okay. We’ve learned to ignore it by now. Just read between the lines for the feel of things and correct it in your mind.

It’s not like we’re cutting-and-pasting these words and attempting to pass them off as human-written.

We get the idea. The concept. The outline. The human can go over things with a fine-tooth comb and write over the skeleton and fix stuff later.

(For instance, in our second pass through this story on the blog, you’ll note I purposefully added yet another nod to this in her origin story itself, creating a prologue scene where she’s an angel archivist before Time began. Just to let it echo fit a little further.)

On the whole, I rather enjoyed myself through this co-written narrative (on one side dice, TYOV prompts and game rules, on another side ChatGPT and in the last corner me), which took about the span of three days for the first pass in a GPT window.

In fact, it had the hallmarks of one of those kinds of dice-led tabletop RPGs, where the dice absolutely know the story they are telling. Somehow.

A completely untrue superstitious belief – it’s more the human pathfinding the way and remembering only those moments that feel mythic – but… still.


Let’s be honest. I didn’t update her character sheet in the actual game once I hit the ending. The story closed. It hit a note I liked. The end.

(Plus a little tiny itch to fix a few narrative inconsistencies and gaps, which we’ll get to in a bit.)

But for this blog, let’s just do it right.

It’s easy enough, despite not really having enough Memory spaces.

All we need to do is put one more Memory into a Diary (it takes up to four) and Memory 5: The Wilds is one of those things she can easily put down and put away.

Just like Wilderness Survival is lost to her, those memories of the wild wanderings are no longer as important to her to keep front and center. Just noted down in written form as a moment in history.

And we start a new Memory 5: The Archive and the Apartment for our ex-angel, now Daoist vampire, in a new modern world, loved and learning how to be.

Really, we lucked into nothing terribly painful on the TYOV memory front.

Astral kept the important stuff she wanted to remember – who she was, people she cared deeply about, stored enough in her head to survive whatever time she was in.

Less important things that could be still relevant or nice-to-keep secondary memories managed to fit into her Diary (probably all verses, knowing her.)

She didn’t encounter any loss of the Diary or damage to it (only starting one after that one prompt) so the memories are all still intact.

The one memory she really lost and mostly forgot was the Salima arc. No great loss there. Too many distasteful reminders of Ishraq and she already did as much right by Salima as she could. The rest – how it turned out after that – is Salima’s own story.

The other memory “lost” was more of a conversion into a Skill. Which -absolutely- played its part throughout the story.

It’s stunningly fitting in the end. Astral tried to look for truth in dusty books and tomes and found only terrible secrets and pain. She tried to deny her vampiric nature through starvation and seclusion, albeit in a place of precious perseverant beauty, and wound up only in torpor.

She converted that memory into an active Skill when she began her Daoist-inflected phase, learning actively and reflecting, on the vagaries of people and how to balance opposing concepts in a kind of harmony. That Skill got her out of being arrested for murder at the hands of a local enforcement lynch mob.

Finally, it converted again into no longer a ‘search for’, but something found.

No regrets with that “lost” memory.

As for the rest? The Marks will always be there. No one gets through life, let alone eternal unlife, without scars.

We squeezed by with no more official Skills on the character sheet. Just. (To think at one point I thought we were accumulating too many Skills.)

Astral’s new skills with a lowercase are more modern ones, I guess. Managing a computer. Library sciences. Actually clothing herself in contemporary fashion.

She still holds one Resource (besides her diary, which technically should be marked on the list, but whatever, I kinda sorta knew it was there.) That esoteric knowledge/spell thing that we started her off with from the beginning of character creation, presumably found in Baghdad’s House of Wisdom somewhere.

I like to think that it was best never used. Perhaps it held the Name of a darker type of fallen angel that she would have done well never to meet or owe favors to.

In any case, I can see her putting that worn and crumbling page carefully away into the hidden archive she now caretakes, filing it under lock and key where it truly belongs, untouched and undiscovered by the wrong hands.

All the Mortals she once knew are dust now. The inevitable fate of the merely human in a story about immortals.

Ishraq may still well be alive. It’s a wide world out there. Too vast for them to ever meet again, let us say.

We know where Iguana is. Always checking in on her angel, if not actually living together for the periods where Iguana doesn’t get restless enough to amble elsewhere and meddle for a while.

Flame probably visits. Now and then. Both finally have those awkward conversations they’ve been putting off for centuries. Both eventually say they’re sorry. They end on an okay note, but live their own lives.

Whether Flame keeps visiting or one day stops popping by, I don’t know. I’ll leave that discovery for if I ever play through TYOV again from Flame’s perspective. There are all kinds of ways for TYOV vampires to meet their end, after all.

For now, this is TYOV Astral’s end.

And the game is over.


Full disclosure: I started smelling the impending ending at around Prompt 66 (Part 23.)

Craving for proper closure of a narrative, what I did was roll the dice first and line up the prompt sequence in that order, to figure out just how many narrative beats I would have left to close the story.

Turns out, I had just three more prompts to go.

So the rest of the story had to be jiggered to fit within those three prompts. If the narrative feels a little short and abrupt, that’s why.

It would have been even more abrupt if we took it prompt by prompt and reacted that way. (Hence the advance dice rolling in an attempt to avoid that.)

Better story structure would mean even more advance rolling work required. Which, kinda goes against the spirit of discovery in a solo roleplaying session, I feel, and would turn it more into a scene outlining and writing exercise.

Not that such a thing is wrong, by any means, and might be fun to do too. Just depends on one’s overall aim when solo roleplaying. How much surprise you want in the process of uncovering the story.

The eternal plotter vs pantser debate in writing, in other words.

Solo RP leans more pantser in nature, and so what we end up with is a rough first cut, the initial draft of a story with potential.

In our second pass through the story via this blog, we used the skeleton of what we already knew to start filling in more of the blanks, craft in more details and callbacks and themes.

It’s always an iterative process.

Amusingly, a TYOV style story spreads out narrative beats so far across time that instead of deleting extraneous scenes, one almost feels obliged to add more scenes, just to fill in the unsatisfactory gaps. Smooth out the abrupt bits, as it were.

And that’s what I started doing with ChatGPT even after the TYOV prompts themselves ended. Just to make the narrative feel a touch better in places I felt were lacking.

Coming up in a few more future posts: bonus scenes and commentary about the whole experience – the second time I played Thousand Year Old Vampire, and really played it to satisfaction, with GPT as assistant GM.

Disclaimer: A Photoshop tweaked AI-generated image was used as the feature image for this post.

Solo RP: Thousand Year Old Vampire – Angel Fall – Part 25

Previously, our vampire lucked into an unexpected reunion, through very little effort of her own and a lot of effort on the other party’s end (which we will cover later)…

Today, our dice roll is a 9 on a 1d10 and a 1 on a 1d6.

Just when I was fearing a lot of slip and sliding back and forth, the opposite happens and we hurtle eight steps forward instead. Help. Too fast now?

Only one Skill left. The hail mary Skill. The last vestige of angelic divinity for our fallen vampire.

In 1960s Japan?

How the hell is that going to work? No clue.

GPT suggestion time please. Link some of those token contexts together and see if you come up with anything that sounds pleasing to me.

Apparently, the phrase “modern world” really dragged GPT into contemporary times.

(I suppose it doesn’t hurt if we want to drag it forward a few decades. Some tens-of-years is a hop compared to the centuries we’ve been jumping, after all.)

However, I feel like the more oldfashioned encounter may be more in order here. At least keep us in the 1970s era, plus minus a decade.

I don’t quite like the idea of a “peaceful” channeled divinity though. The Prompt wants something costly. It needs to hurt a bit more. There’s one other option that suggests an idea along those veins.

As always, the human steers. The AI doesn’t actually know what it’s doing.

Let’s exercise the out-of-the-box option and pick a little from column A and a little from column B.


That Experience fills in the last slot in Memory 4: Changing Times.

No Skills left. One final Resource.

On a meta level, just flipping through the PDF, it’s very obvious we’re nearing the end of the TYOV prompts.

The only real question is, which final prompt are we landing on, and will the dice make us do more back-and-forth until we run out of Resource and meet a bad end regardless?

Disclaimer: A Photoshop tweaked AI-generated image was used as the feature image for this post.

Solo RP: Thousand Year Old Vampire – Angel Fall – Part 24

Previously, our vampire found the entire world changing on her as time inexorably moved forward, while we had a suspicious feeling that the end is nigh…

Our next roll is a 2 on a d10 and a 6 on a 1d6.

Negative number, here we go, sliding back again like snakes and ladders:

Ahh.

Something that makes my vampire’s life more satisfying.

An ancient taboo lifted, if only in Astral’s own silly little head.

I have been sitting around, waiting and praying for a Prompt that calls in an Immortal. Have not encountered one in forever.

But this? This is the perfect excuse for an angel of narrative’s deus ex machina.

And all it costs is changing a Skill. Could be checked or unchecked. That’s like practically free.

Dice gods for the win.

Let’s not stop at one retelling of this moment. There are always many possible ways to tell the same story. GPT can show us variations to consider, after all.

All that lost wandering over centuries… then finally found. Needs at least one more look.

Pretty sure streaming music didn’t exist in the 1960s, GPT.

(On the other hand, -I- Google researched that bit about CCTVs, enough to decide it’s plausible enough to have public surveillance in Japanese parks in the 1960s-1970s. For a fictional self-told story, anyway.)

It’s okay. We’ll forgive a few anachronisms.

The mood and feel of my original scene skeleton was captured. Good enough.

If we ever do a human rewrite of this, we’ll fix it. That’s what iterated drafts are for.

We get a little carried away and try another sparse prompt GPT test. It offered a first night back together scene.

How much has it understood from context? And how much will it screw up?

*twitches* First off, vampires don’t drink tea. Not in my book anyway. Which lax vampire stories has GPT been reading?

Then a really weird line involving my other Ares characters snuck in – possibly because I locked them in GPT’s memory – but it doesn’t quite fit into an AU where I never brought them in.

Yea, well, it’s not great. Can’t win ’em all. Better to steer GPT where you want. Getting it to make stuff up is… just random possibilities. Might hit upon a few nice things faster than monkeys on typewriters, if you’re lucky. If not, it’s just… words. Some acceptable. Some not.


Possibly the most satisfying change / loss of a Skill ever. I love the poetry in changing something Searching for Truth into someone having Found a Truth.

Not THE Truth, of course. Don’t think there’s such a thing. Just A truth. Something that works and rings true to that person.

And this whole damn narrative has been Astrael looking for her goddamn exiled angel and getting lost along the way. (Reaching some other truth in the process, yes, just not the original one.)

Turns out, someone’s also been looking.

The Experience slides neatly into the new Memory 4: Changing Times as 4b.

It’s a new world, but at least our vampire doesn’t have to deal with it alone any longer.

Mind you, we are still dancing on the edge of disaster here with one Skill and one Resource left. How many times will the dice send us back and forth? And what kind of end is it sending us forward into?

Astral’s new modern living is still not all sunshine and roses, as our next Prompt will attest

Disclaimer: A Photoshop tweaked AI-generated image was used as the feature image for this post.

Solo RP: Thousand Year Old Vampire – Angel Fall – Part 23

Previously, our vampire had dogs unleashed on her by unfriendly farmers

Today we roll a 5 on a 1d10 and a 2 on a 1d6, skipping forward three paces into:

Oh my.

I have two unchecked Skills left. Channel Divinity and (darn it) Wilderness Survival.

The former has been hoarded through so many prompts as my emergency hail mary, and it also doesn’t sound like knowledge or something that can be outmoded.

Which leaves (aaargh) Wilderness Survival. Something so core to our vampire’s survival strategy for centuries that it hurts to lose it now.

But narratively, it also fits so well.

All we need to do is skip forward further into time, don’t we? Reach forward towards modernity, towards a time when the Silk Road becomes history.

And this is an example of ChatGPT being a ridiculously good ideas partner when you get lucky and somehow prompt it just right.

Could not have done better. Had the idea, had the vibe and GPT narrated the rest and the passage of time.

Not much more to fill in, really. Our vampire keeps drifting and is steadily getting outpaced by modernity, her love of the shrinking wild places becoming less viable – she still needs to be where the blood is – and her style of survival becomes more of a sideshow, an oddity.


New Experience, representing a really long time period. Out of space and definitely needs a new Memory for this one.

So we stuff Memory 4: Mortal Fear into a Diary. The sequel story of Chimeko and Tenzin is remembered in written form, probably in verse again. But we’ve skipped forward centuries since. All the mortals she knew are long dead. We cancel them all out too.

New Memory 4: Changing Times. (Oh the times, they are a-changin’)

Wilderness Survival? Gone. From the whole stat block. Hurts so much.

One Skill left. One Resource left.

We can go through one Prompt still, and the Prompt after that is going to demand a cost if we need to convert an inappropriate Skill/Resource to the other form, and the next Prompt after may mean the end. In a bad way, if she doesn’t have Skill or Resource left. Getting a little tense here.

Just going to have to see what happens next and keep rolling those dice


Disclaimer: Photoshop tweaked AI-generated images were used as feature images for this post.