So basically, I found myself in the situation where the best way to make some sort of tutorial for my game is basically an introductive video, with audio, and I have no way to get someone to voice the video (nor I want or can use my voice (it isn’t even the type of voice I need)) so I wanted y’all’s opinion on using an AI voice.
Makes total sense, don’t know if I’d have any way to put it in a senseful way considering the game’s context, even though an idea just popped onto my mind.
That is a tough one. Try it on a free service and see how you feel about it. Depending on how much talking you need, you could try and get a cheap service on fiverr perhaps? Although from experience you will end up needing to try out more than one service before you get something useable.
As for your own voice, record it yourself as best you can and then use an audio tool for pitch shifting, synth effects etc. Acting itself is quite hard but you should be able to produce something that does not sound like you, but is you, without using AI voices. I would try that first before paying for it.
AI voice tools are getting better, in fact, in some circumstance are quite amazing. I would not discount using it personally, but as @that_duck rightly pointed out, the “uncanny valley” from AI voices means you would have to quality control yourself and be very aware of. Also it can turn alot of people off immediately if they recognise the voice as AI.
Would you let us know how it went and what you did? I want to add voices to my game (not yet but at some point) but have not got there yet.
This biggest technical problem using generative AI for voices is conveying the appropriate emotion. I have found that while you can do a lot if you REALLY know how to write well, it’s still hard to direct. That means correct spelling, as well as knowing how to use commas, semicolons, hyphens, ellipses, and . . . when to spell words incorrectly to get the pronunciation you want.
If you want something real time because you want to also generate what NPCs are saying, you can check out the free tier of Amazon Lex or Murf AI.
If you want to just take some dialogue and do a once over, there’s a few options to try (last I checked.) There’s Luv Voice which I wasn’t too impressed with. There’s Text To Speech Maker which did much better, but I ultimately abandoned in favor of my own voice recording because I couldn’t get the narration voice I wanted. Finally, there’s Eleven Labs which was recommended to me and looks really promising, but I haven’t actually tried yet.
Another possible option is using a music creator like Suno. You could tell it you want narration over a musical background. You could even upload the music you want to use (as long as you have the permission of whoever made it) and use Suno to add “Vocals”.
While all these options have free entry tiers, you may end up either having to go with “good enough” or paying money to play around enough or to have access to all the tools of a suite.
There are also voice replacement AI tools. I’ve seen them used for music. You feed it the voice line of a song and select a new voice to sing it. You could record your own voice saying things and then use AI voice replacement and that would get you the right inflections.
The one issue I have with Ai voices is when there’s no human performance driving it. However, there’s this platform called eleven labs, it actually lets you record the performance with your voice and keeps your “acting” and emotion. It just changes your voice.
I’ve never used it, just saw a video, so you might want to try it out.
Can I ask for more info on the situation? I’m interested.
I know Nintendo often uses videos to show how to move their controllers, so that’s a reason I can think of, custom controllers.
Anyway, there’s nuance in the use of AI for voice acting.
I personally opt for not using it, and disagree with most cases. Robotic AI voices in smaller, especially meme-like, games, I don’t mind. It can be a part of the charm, but I also would see it as less professional without story relevance or something.
As a final note, there are some online willing to voice act. I know of TheIndieManiac on Twitter; he’s got connections and does some voice acting himself. I can also say subtitles and small sound effects can go a long way for indie titles.
It’s all about making it not feel off, though, so QA and all that.
I think AI voice acting could be an invaluable tool for indy creators who lack the funding to hire real actors.
If you wanted to translate your game to other languages, you would have to select one that could also speak those languages. I would expect google might have one, and again this option seems cheaper especially for larger projects.
That is true. You also can’t really do this very easily without paying one of them anyway. But yes, if you go the Suno route you need to pay. In my opinion it’s worth the money. The tools get a lot better with the first tier alone.
That’s what I did (orbitrel.com) and I’m satisfied. It’s imperfect but considerably better than my own voice, and hiring an actor makes no sense at this scale.
Looks like Abode just a released a tool too. It does music and voiceovers, and allows you to correct pronunciation and tone in speech. What I found interesting is they made sure to only train their tool on music they got rights to train it on. Which gets rid of some ethical issues, if not others.
AI voice is really not worth the reputational risk ngl, there are a bunch of folks who will happily do cheap or free voice acting for projects online but if you have that AI content disclosure on your steam page your playership will fucking crater
As much as I vividly hate generative AI, I can’t help agree with @dragonforge-dev.
AI hasn’t existed for long enough for anybody to have any valid long-term data. It’s actually insane, because AI in its current form has only been around for 3 years! ONLY THREE!!! And it’s already changed so much.
Also, since generative AI is still so young, the (slow AF) global legal systems has yet to decide how it’s regulated, how the tech unfortunately effects our lives, and how the economic AI bubble propped up by investors with the promise of AGI will pop. And I believe it WILL pop sometime in the next few years.
ChatGPT/OpenAI is RIDICULOUSLY unprofitable and the magic is starting to wear off for some of the general populous. None of us know if it will even exist in 5 years. The general tech will stick around, but not in this wild west/arms race type deal.
As for AI voice “acting”, it’s a gamble. Do it at your own risk. I’m not talking about backlash from butthurt people like me, I mean the legal side of things.
If you turn out just fine, I’ll happy for you. But I wouldn’t take the risk right now for any long-term/serious project.
Edit:
I also wanted to add, if you DO use AI, make sure to do it well. Because poor quality anything will be noticed by players. But bad quality AI will have an even worse response, because it gives the impression you didn’t put work into the project. (Most poor quality human-made content had someone honestly try their best)
I’m not saying you won’t put in effort, quite the contrary, but impressions matter a TON. AI tends to be used to quickly make cash-grab shovelware games on Steam and slop on the internet at large (Have you SEEN Sora?!). You’ll have to be on your A-game if you want to fight against that stigma.