AI Chat

[Booker Grimm]

Some of the other VTTs are already taking advantage of AI chat to voice thier NPCs.

What is the plan for FGU?

[Zacchaeus]

I haven't heard of any plans to introduce that. What use case would there be for it? What sort of stuff would that bring to the table?

[LordEntrails]

I suspect using AI (or really, it's just text to speech) for voicing NPCs is going to best be handled by a dedicated program, like Dragon, Clownfish or something new (I'm not real informed on that market). Sure, in some years when such programs are mature then I can see an interface like we now have with Syrinscape. But I would hate to see the devs that already have a huge list of feature requests spending hundreds of hours incorporating something like NPC voices.

[Vass_Dts]

I suspect using AI (or really, it's just text to speech) for voicing NPCs is going to best be handled by a dedicated program, like Dragon, Clownfish or something new (I'm not real informed on that market). Sure, in some years when such programs are mature then I can see an interface like we now have with Syrinscape. But I would hate to see the devs that already have a huge list of feature requests spending hundreds of hours incorporating something like NPC voices.

Plus, if a GM really cares for voice alteration, there are already apps/services that do that in real time. I personally use voicemod to change my voice for specific NPCs.

A full AI integration would enable NPCs that reply on their own without GM input, which is something I personally don't care at all about. If automated spell animations push a VTT to full video game territory on a grade of say 3, then AI would push that to 11. That's not my cup of tea.

[DaddieCat]

I would also second Voicemod or just plain making voices, silly or otherwise.

[Booker Grimm]

I tried a test with Chat GTP open and gave it a few lines about an NPC. It voiced the results using a different voice for Narrator and NPC. The following was all generated on just these few spoken lines.

Colt is a male cultist. He's arrogant, dangerous but above all self preserving. He taunts the PCs in their fight, threatening them. Make up a few lines for something he might say while fighting the heroes.

[Remember Colt's characteristics and come up with some lines for him. The party are engaged in a fight and now it's Colt's turn. And he has already been wounded by Amber, the party's female leader.

ChatGPT said:
ChatGPT

Colt staggers back, clutching the wound Amber inflicted, his eyes burning with a mix of rage and arrogance. He snarls, sneering at her through gritted teeth.

In a voice dripping with contempt:
"Amber, is it? I’ll make sure your name is the first to be erased when I’m done with you."

He glances at his bleeding wound, his smirk returning.
"You think this? This is enough to stop me? I’ve bled before—far worse than this—and I still stand. You? You’ll crawl."

Colt raises his weapon, taunting Amber, his voice dropping to a venomous whisper:
"I’ll carve you apart piece by piece, and when you beg for mercy, I'll laugh."

He shifts his gaze to the rest of the party, voice rising in cruel mockery:
"You’re following her? How sweet. Watch her fall, and then maybe you’ll understand the futility of defying me."

He lunges forward, reckless but dangerous, driven by his desperation to survive.

Hearing that and have the lines generated by just a few simple phrases that can be set up before hand and having them in the game is really amazing. An gives the DM great ideas even if he doesn't use them. I note that Fabled (and some of the more recent VTTs are going to include AIs as standard. I think FGU should get onboard.

[Laerun]

I tried a test with Chat GTP open and gave it a few lines about an NPC. It voiced the results using a different voice for Narrator and NPC. The following was all generated on just these few spoken lines.

Colt is a male cultist. He's arrogant, dangerous but above all self preserving. He taunts the PCs in their fight, threatening them. Make up a few lines for something he might say while fighting the heroes.

[Remember Colt's characteristics and come up with some lines for him. The party are engaged in a fight and now it's Colt's turn. And he has already been wounded by Amber, the party's female leader.

ChatGPT said:
ChatGPT

Colt staggers back, clutching the wound Amber inflicted, his eyes burning with a mix of rage and arrogance. He snarls, sneering at her through gritted teeth.

In a voice dripping with contempt:
"Amber, is it? I’ll make sure your name is the first to be erased when I’m done with you."

He glances at his bleeding wound, his smirk returning.
"You think this? This is enough to stop me? I’ve bled before—far worse than this—and I still stand. You? You’ll crawl."

Colt raises his weapon, taunting Amber, his voice dropping to a venomous whisper:
"I’ll carve you apart piece by piece, and when you beg for mercy, I'll laugh."

He shifts his gaze to the rest of the party, voice rising in cruel mockery:
"You’re following her? How sweet. Watch her fall, and then maybe you’ll understand the futility of defying me."

He lunges forward, reckless but dangerous, driven by his desperation to survive.

Hearing that and have the lines generated by just a few simple phrases that can be set up before hand and having them in the game is really amazing. An gives the DM great ideas even if he doesn't use them. I note that Fabled (and some of the more recent VTTs are going to include AIs as standard. I think FGU should get onboard.

Most of these tools incur third party costs. There are text to voice AI apps already, but they usually have limits in Free mode. If you had an example of the VTT that uses a built in tool and such it'd be interesting to see or hear in action.
A one off situation works well, but any form of consistency is still difficult to achieve due to context, token length, and where AI tools are at currently. Recall is tough. Eventually an AI generated NPC voice dialog will hallucinate or drift and forget the previous conversations.
Until consistency, stability, and more voice type selections are available, it does not seem viable enough to use currently without a bunch of work, time, and potentially cost.
We will have to see what the future brings.

[Zacchaeus]

How did you set this up and what software is needed to play these voices and how did you pipe that to the players?