To be fair, it was canonically an unauthorized deepfake! Not that I have anything against Hertzler, but just sayin’ that inauthentic is authentic in this particular case.
Plus, I’m talking about using an LLM with infinite possible responses, not something akin to the TNG interactive VCR board game from the '90s. At best, it’s still just a text-to-speech trained on his voice with his permission instead of without it, not actually him voice-acting each line.
I get you. Given enough time, this type of notion will warrant reconsideration and possibly result in worthwhile media. We’re just not there presently.
You can do that now. That’s really the only thing I’ve done with a chatbot I thought worked out pretty realistically; using one that lets you get multiple bots in the same chat and playing D&D with them.
The inconsistencies of the output worked in its favor in-so-far as you’ll spend half the session arguing over the interpretation of the rules and forgetting what happened previously. Just like at a real table.
The chatbot in question was already designed to be for fictional characters. Getting Martok should be easy.
I know AI isn’t popular these days, especially here on Lemmy… buuut…
How far off are we, actually, from being able to have an unauthorized AI Martok DM in real life?
It ain’t worth it if it’s not authentic. J.G. Hertzler’s still above ground, and if Tawny Newsome has some pull accompanied by the fanbase…
To be fair, it was canonically an unauthorized deepfake! Not that I have anything against Hertzler, but just sayin’ that inauthentic is authentic in this particular case.
Plus, I’m talking about using an LLM with infinite possible responses, not something akin to the TNG interactive VCR board game from the '90s. At best, it’s still just a text-to-speech trained on his voice with his permission instead of without it, not actually him voice-acting each line.
I get you. Given enough time, this type of notion will warrant reconsideration and possibly result in worthwhile media. We’re just not there presently.
You can do that now. That’s really the only thing I’ve done with a chatbot I thought worked out pretty realistically; using one that lets you get multiple bots in the same chat and playing D&D with them.
The inconsistencies of the output worked in its favor in-so-far as you’ll spend half the session arguing over the interpretation of the rules and forgetting what happened previously. Just like at a real table.
The chatbot in question was already designed to be for fictional characters. Getting Martok should be easy.
But how close are we to getting the response as a deepfaked video in real-time, instead of text?
It would be pretty trivial to create a martok voice clone and pipe llm text output to a TTS model with martok’s voice.
The video portion idk about but it’s probably doable.
Pretty sure we’re there with that, too, at this point. Sora was already really fuckin’ hard to tell it was AI.