AI Chat

[LordEntrails]

This is much more promising than I was expecting.
Is it possible to save personalities between game sessions though?

[Booker Grimm]

This is much more promising than I was expecting.
Is it possible to save personalities between game sessions though?

Yes. It remembers every chat. The free version is text only. You have to subscribe for full audio, but so far I am very impressed.

[LordEntrails]

Would love to see a video of you using it like this. The text descriptions are nice, but still leave some questions.
But, if you're not setup for that, it's ok, I hope to be able to try this myself in the next few weeks.

[Booker Grimm]

Would love to see a video of you using it like this. The text descriptions are nice, but still leave some questions.
But, if you're not setup for that, it's ok, I hope to be able to try this myself in the next few weeks.

Have a look at any decent chat AI and play around with it. I used Chat GTP because it was free to register. The desktop version has what you need to test encounters and roleplaying too.

Here's another example of an encounter with some thugs. Again I told it we are playing 5e and this time to include attacks and damage.

You said: 5 thugs, armed with crossbows and scimitars attack the party. Sar Bonecarver is the party leader and has an AC of 14, Alexi is his friend and has an AC of 19. It's the thugs turn.


Chat GPT:
Five thugs, their faces obscured by shadowy hoods, surround the party, armed with crossbows and scimitars. They move in sync, calculating their strikes as they aim to overwhelm their targets.

A thug with a scar across his cheek raises his crossbow, aiming directly at Sar Bonecarver. He pulls the trigger with a sneer.
(Attack roll: [d20 + 4] = 10)
The bolt flies wide, missing Sar completely.
"You’re quicker than you look, old man!" he growls, frustration creeping into his voice.

Another thug, with a chipped scimitar glinting in the dim light, lunges at Alexi with a wicked grin.
(Attack roll: [d20 + 4] = 14)
His blade bounces off Alexi’s armor with a clang, the strike ineffectual.
"What? Is that all you’ve got?" Alexi taunts, deflecting the blow.

A third thug, slightly larger than the rest, fires his crossbow at Sar with a determined look.
(Attack roll: [d20 + 4] = 16)
The bolt strikes true, embedding itself in Sar’s shoulder.
(Damage: [1d6 + 2] = 4 piercing damage)
"Take that, Bonecarver!" he shouts, his voice filled with glee.

The fourth thug, eager to join the fray, charges in with his scimitar drawn, aiming for Sar as well.
(Attack roll: [d20 + 4] = 8)
His swing goes wide, barely missing Sar as he dodges out of the way.
"You can’t run forever!" he yells, panting from the effort.

Finally, the last thug, a wiry figure, steadies his aim on Alexi with a wicked smile.
(Attack roll: [d20 + 4] = 19)
The bolt finds its mark, striking Alexi hard in the side.
(Damage: [1d6 + 2] = 6 piercing damage)
"That’ll teach you to mess with us!" he cackles, reveling in the moment.

This was all read by the narrator (which is an American Female).

I'll try and get a recording going soon.

[Laerun]

Have a look at any decent chat AI and play around with it. I used Chat GTP because it was free to register. The desktop version has what you need to test encounters and roleplaying too.

Here's another example of an encounter with some thugs. Again I told it we are playing 5e and this time to include attacks and damage.

You said: 5 thugs, armed with crossbows and scimitars attack the party. Sar Bonecarver is the party leader and has an AC of 14, Alexi is his friend and has an AC of 19. It's the thugs turn.


Chat GPT:
Five thugs, their faces obscured by shadowy hoods, surround the party, armed with crossbows and scimitars. They move in sync, calculating their strikes as they aim to overwhelm their targets.

A thug with a scar across his cheek raises his crossbow, aiming directly at Sar Bonecarver. He pulls the trigger with a sneer.
(Attack roll: [d20 + 4] = 10)
The bolt flies wide, missing Sar completely.
"You’re quicker than you look, old man!" he growls, frustration creeping into his voice.

Another thug, with a chipped scimitar glinting in the dim light, lunges at Alexi with a wicked grin.
(Attack roll: [d20 + 4] = 14)
His blade bounces off Alexi’s armor with a clang, the strike ineffectual.
"What? Is that all you’ve got?" Alexi taunts, deflecting the blow.

A third thug, slightly larger than the rest, fires his crossbow at Sar with a determined look.
(Attack roll: [d20 + 4] = 16)
The bolt strikes true, embedding itself in Sar’s shoulder.
(Damage: [1d6 + 2] = 4 piercing damage)
"Take that, Bonecarver!" he shouts, his voice filled with glee.

The fourth thug, eager to join the fray, charges in with his scimitar drawn, aiming for Sar as well.
(Attack roll: [d20 + 4] = 8)
His swing goes wide, barely missing Sar as he dodges out of the way.
"You can’t run forever!" he yells, panting from the effort.

Finally, the last thug, a wiry figure, steadies his aim on Alexi with a wicked smile.
(Attack roll: [d20 + 4] = 19)
The bolt finds its mark, striking Alexi hard in the side.
(Damage: [1d6 + 2] = 6 piercing damage)
"That’ll teach you to mess with us!" he cackles, reveling in the moment.

This was all read by the narrator (which is an American Female).

I'll try and get a recording going soon.

The GPT models have a token length limit. After awhile, the context gets lost, buried, and irrelevant. Conversations hit a wall after awhile. Also, the memory is not always accessed or used in the way one would hope, it's very hit and miss. The token length has increased over the months, but it is still not quite there yet. Anthropic's Sonnet does a lot better with this, but still has similar context length limits and no images. Google's Notebook LM is pretty good for keeping notes, world building, and creating content based upon what documents, links, and text that you feed it, but it too has limitations and is in Beta still.

When you're deep into a long RPG chat, like designing or playing through a campaign, running into token limits can disrupt the flow. Even though this version of GPT-4 has a much larger context length (128k tokens), which is roughly equivalent to 96,000 words, it will eventually hit a wall after extensive back-and-forth.

When that happens, older parts of the conversation start getting pushed out of memory, which can cause some of the earlier details to be forgotten or misunderstood. If you're building a complex world or campaign, this can mean needing to repeat certain aspects or clarifying things you've already mentioned.

To potentially help avoid that in long RPG sessions:

1. Summarize Key Points: We can summarize major plot points, character details, or rules periodically. This way, I keep track of what’s essential without holding onto the entire conversation.


2. Segment Conversations: If we break the game into smaller arcs or sessions, each with its own set of important details, we can focus on specific areas without overwhelming the token limit.


3. Keep an External Log: You can export summaries or logs to refer back to when needed, reducing reliance on in-chat memory.

[claedawg]

One of my players recently took my player campaign document and loaded into Google's Notebook LM to create a podcast. It was pretty cool.
https://notebooklm.google.com/notebo...68/audio?pli=1