Voice Acting by AI
Posted: Tue Dec 19, 2023 12:29 pm
One of the things I'm keen to add into my current project (and future projects, hopefully) is voice acting. However, I'm not currently in a position to commission a VA to speak the lines of one character, let alone all of them. Initially, I looked into self-voicing, but it appears as though that would use the system default voice for everything, with no option to switch voices on the fly, depending on which character is speaking.
As an experiment, I recorded myself as the 'narrator', just to get an idea of how to add voice tracks, but my voice just doesn't suit the female main character's internal dialogue. It was a fun process, certainly, but not much practical use. Over the weekend, I started looking into AI-generated voices as a stopgap, using Murf.ai, and now have a full set of narration and dialogue for my current work-in-progress demo.
It's... serviceable. The rhythm is better on some lines than others, the pronunciation relies on either choosing an established option (some of which just don't sound right to my ears) or attempting to decipher the custom pronunciation options, while adding emphasis became a real chore that seldom resulted in precisely what I was after. I certainly couldn't use it as a guide to direct a human VA, let alone in a 'finished product' VN, but it's 'good enough' as a proof of concept.
Has anyone else tried using AI to create voice tracks on a zero budget? If so, what were the results like... If not, why not?
What's the general consensus (if any!) on AI-generated voices, given that AI-generated art is widely considered pretty loathsome?
As an experiment, I recorded myself as the 'narrator', just to get an idea of how to add voice tracks, but my voice just doesn't suit the female main character's internal dialogue. It was a fun process, certainly, but not much practical use. Over the weekend, I started looking into AI-generated voices as a stopgap, using Murf.ai, and now have a full set of narration and dialogue for my current work-in-progress demo.
It's... serviceable. The rhythm is better on some lines than others, the pronunciation relies on either choosing an established option (some of which just don't sound right to my ears) or attempting to decipher the custom pronunciation options, while adding emphasis became a real chore that seldom resulted in precisely what I was after. I certainly couldn't use it as a guide to direct a human VA, let alone in a 'finished product' VN, but it's 'good enough' as a proof of concept.
Has anyone else tried using AI to create voice tracks on a zero budget? If so, what were the results like... If not, why not?
What's the general consensus (if any!) on AI-generated voices, given that AI-generated art is widely considered pretty loathsome?