Honestly, I think we're almost there. The example you showed with your voice is so incredibly close, and this is really just about having enough data to sufficiently emulate not just a voice, but also a style and an approach. "Voice" has many meanings when it comes to text, and I think this is where the breakthrough will be. If ChatGPT gets better at mimicking my tone and audio AI gets better at synthesizing my sound, then I am quite afraid for my "job." The interesting part will be proof of humanity. Right now, we're looking at how we can determine if something is made by AI. But at the rate of progress, we should really be finding a reliable proof of when something is made by humans.
The interesting question to me is whether an AI can sustain this kind of output over a 30-60 minute episode. The synthetic voices by ElevenLabs do a pretty good job with 1-2 minute clips, but not sure how a whole episode would sound.
I could have been fooled by Justin's AI replicant - to believe it's Justin, but a more boring Justin. The chilling thought is not whether AI can replace Justin, but what if we found out the world is satisfied with a slightly more boring version of us all. Let's give it a fight and be less predicatble! Let's revive the Dada art genre, from the previous century!
Interesting video! The speed at which AI is evolving is truly impressive, and it's intriguing to think about the potential impact on the future of podcasting. I absolutely prefer a human, real, voice, but we are very close to a point where it won't be possible to know if a voice is real or not. Maybe in a few months, a year, or two years we probably won't be able to distinguish what's AI generated and what's not.
If anyone wants to clone their own voice, you can use this tool from ElevenLabs (no coding required): elevenlabs.io/voice-cloning
Honestly, I think we're almost there. The example you showed with your voice is so incredibly close, and this is really just about having enough data to sufficiently emulate not just a voice, but also a style and an approach. "Voice" has many meanings when it comes to text, and I think this is where the breakthrough will be. If ChatGPT gets better at mimicking my tone and audio AI gets better at synthesizing my sound, then I am quite afraid for my "job."
The interesting part will be proof of humanity. Right now, we're looking at how we can determine if something is made by AI. But at the rate of progress, we should really be finding a reliable proof of when something is made by humans.
The interesting question to me is whether an AI can sustain this kind of output over a 30-60 minute episode. The synthetic voices by ElevenLabs do a pretty good job with 1-2 minute clips, but not sure how a whole episode would sound.
@@TransistorPodcasting Soon to come: the AI that blends 2min clips into a larger narrative :D
@@TheBootstrappedFounder 😱
I could have been fooled by Justin's AI replicant - to believe it's Justin, but a more boring Justin. The chilling thought is not whether AI can replace Justin, but what if we found out the world is satisfied with a slightly more boring version of us all. Let's give it a fight and be less predicatble! Let's revive the Dada art genre, from the previous century!
I think the example with your voice, I could not tell if it was really you or AI 🤯
Sigh. *signs up for ElevenLabs* 😂
Interesting video! The speed at which AI is evolving is truly impressive, and it's intriguing to think about the potential impact on the future of podcasting. I absolutely prefer a human, real, voice, but we are very close to a point where it won't be possible to know if a voice is real or not. Maybe in a few months, a year, or two years we probably won't be able to distinguish what's AI generated and what's not.