3 ms·

I suppose you're paying a lot for the voice cloning, so do consider that in many countries voice-overs are done with a single generic voice. Would you consider

by exitb 3y ago

I suppose you're paying a lot for the voice cloning, so do consider that in many countries voice-overs are done with a single generic voice. Would you consider a lower price point service doing just that? I'm doing something similar and using GPT-4 for translation. What's unique about it, is that you can specifically prompt it to avoid long translations by rephrasing things, so you can buy yourself some time for the "Fledermaus".

leobg 3y ago
Using a single speaker makes sense when you're paying for human voice talents. But if you're using a computer to generate the voice, why not generate a voice that sounds like the original speaker? Much more fun to hear Elon speak Chinese :)
- exitb 3y ago
  To be blunt, because of the price. I’m running the whole pipeline of my toy project for much less than $5 per hour. If the voice cloning is the long pole in the tent, I’d just consider dropping it. Moreover, it’s a cultural custom. I’m from Poland and here the voice-over narrator is supposed to be generic and bland, so your brain learns to tune him out and take the emotional cues from the original voice.
  - leobg 3y ago
    Yes, in Germany we have generic voice-over narrators, too. In documentaries, etc.. They usually match the gender of the speaker, but that's it. Personally, I read most of my books with the iOS app Voice Dream Reader. That app still uses old TTS voices. They sounded great 3 years ago, but now sound robotic when compared to Elevenlabs or WaveNet. But, as you say, you learn to tune out the voice. I can read entire novels like this, and I still "hear" the different voices and personalities. It just all happens in my imagination. How much I'd need to charge to make the project worthwhile depends on many factors. And I didn't want to name a price now and then backpaddle a month from now and say it'll actually cost more. My pipeline right now is super unoptimized, to the point of being embarrassing. This can all be made to run much faster and cheaper. I agree with you that if the voice cloning part of my pipeline causes a significant chunk of the cost that the end user pays for the service, that I should then offer the option of using a "bland" voice instead for a lower price.

by exitb 3y ago

leobg 3y ago

Using a single speaker makes sense when you're paying for human voice talents. But if you're using a computer to generate the voice, why not generate a voice that sounds like the original speaker? Much more fun to hear Elon speak Chinese :)

exitb 3y ago
To be blunt, because of the price. I’m running the whole pipeline of my toy project for much less than $5 per hour. If the voice cloning is the long pole in the tent, I’d just consider dropping it. Moreover, it’s a cultural custom. I’m from Poland and here the voice-over narrator is supposed to be generic and bland, so your brain learns to tune him out and take the emotional cues from the original voice.
- leobg 3y ago
  Yes, in Germany we have generic voice-over narrators, too. In documentaries, etc.. They usually match the gender of the speaker, but that's it. Personally, I read most of my books with the iOS app Voice Dream Reader. That app still uses old TTS voices. They sounded great 3 years ago, but now sound robotic when compared to Elevenlabs or WaveNet. But, as you say, you learn to tune out the voice. I can read entire novels like this, and I still "hear" the different voices and personalities. It just all happens in my imagination. How much I'd need to charge to make the project worthwhile depends on many factors. And I didn't want to name a price now and then backpaddle a month from now and say it'll actually cost more. My pipeline right now is super unoptimized, to the point of being embarrassing. This can all be made to run much faster and cheaper. I agree with you that if the voice cloning part of my pipeline causes a significant chunk of the cost that the end user pays for the service, that I should then offer the option of using a "bland" voice instead for a lower price.