Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines

We used AI to voice characters in our visual novel. What do you think of this?

A topic by Carnivore Mors Interactive created Jan 06, 2023 Views: 2,158 Replies: 29
Viewing posts 1 to 14
(+1)

Hello, everyone!

Carnivore Mors Interactive team here. We wanted to know what you thought about AI voiceover. It is not an easy technology to use and achieve good results with, but it still works. We used Replica AI to provide voices for characters in our visual novel. Here's a video example (I hope this doesn't violate the no self-promotion rule).

"The Day That Changed My Life" Edetta and Gia Full Scene - YouTube

Do you think it sounds good to you? Do you believe AI voiceover is preferable to no voiceover (including the option to disable it)? Should the game notify the player that an AI voiceover is being used?

(+1)

I  checked the first two minutes and didn"t hear any voices...

(+1)

Are you sure? Check out 0.54 and 1.16 :)

(3 edits) (+2)

I don’t see a reason to review it. I already know never to play your games.

Are you sure you’re a team, or are you one guy who’s automated everyone else away? If not, when do you project you will?

(+2)

Oh wow! What happened? Did my inquiry offend you? We are a four-person team working hard on the project.

(+1)

Yeah, I can tell. And tell me this: when AI improves to the point your remaining three teammates become expendable, will you throw them in the trash as well? Why should they trust you?

(+3)

No AI will ever replace anyone on our team because ideas in our team are generated through discussion, which no AI can do. I even wrote a post about it once. There is no need to react negatively to an attempt by a team working on ethusiasm only to create a voiceover for their project, especially since it may assist visually impaired people to play it. Chill. I respect and agree with your viewpoints, but this is not the case when someone attempts to profit from AI replacing human labour.

(1 edit) (+1)

The free market is value-neutral. It is easy to blame an inanimate object, but vice was ultimately your choice.

(+3)

This sounds like a sermon delivered by a priest. Is there an anti-AI church? If replacing "AI" with "sampled voices" in the post helps, please read it again. Regarding AI, I've taken note of your thoughts; your perspective is appreciated. I really, really, really hope you can calm down and see the world and the events in it with a more optimistic perspective.

(+3)

Lol dude you use free assets in your games and then complain about people using a tool to do voiceovers.

(2 edits) (+1)

Personally I have my reservations about these technologies, but I will not go into details, since it does not apply and I will focus on answering your question.

The dialogues are too short to be able to draw conclusions, the problem with computer generated voices is the acting, usually the intonations do not convey the correct feeling.

3 or 5 lines of a few words, are too few examples to realize if the performance enriches the dialogues or not. The best thing is that instead of a 4 minute video with only a few seconds of voice, you create a video dedicated solely to voice narration, preferably in different situations.

Currently, what you have doesn't seem bad, but I really wouldn't play it, because starting to hear the voice, then stopping after the second or third word, has always bothered me, you don't know if it's a mistake, if you should read, or if you must listen and in that process, you lose focus on the game and the story.


EDIT:
"We wanted to know what you thought about AI voiceover"
The previous person has given you feedback, and instead of analyzing the actual effect it might have on a target audience, you go on the offensive.

My advice, if you are looking for opinions, you have to learn to read the good and bad opinions and learn from them.

I recommend that you give a read to this thread that talks a bit about how to "filter" the feedback.
https://itch.io/t/2570482/gamedev-tips-for-2023


EDIT2:

It does not matter if the voice is generated by a computer or if it is a real actor, it is best that you always give the option to turn them off.

Personally, I would like the game to indicate if it uses an AI generated voice, but this is a personal opinion, I couldn't tell you if objectively, it is better or worse.

(+1)

Hello there!

Thank you very much for your contribution! I agree that hearing a couple of lines makes it difficult to judge overall quality. We actually have a completed scene that is entirely dialogue, and the synthesised voiceover is extremely difficult to perceive and destroys the atmosphere. We decided to include such a voiceover because I know there are visually impaired people who rely on screen narrators to read for them, and I believe that adding different voices would provide a much better experience for them. It is obviously only an option; it is not even enabled by default. It's a visual novel, so you should read it.

I disagree with you about my being offensive. I mentioned that I had heard and respected a viewpoint expressed by "mid", but this person did nothing but blame me for things I had not done. So I believe that being a fluffy snowflake is not the appropriate reaction in this case, and I reiterate that I heard and support the opinion. Anyway, thank you for the link to the post you provided; it was very refreshing.

Concerning the indication that the Voiceover is synthesised, I agree with you, and I believe we will definitely include this in the description of the "voiceover" option.

Thank you so much for your detailed response!

I've listened to the voice from the video and although it does sound good as expected, the intonations are wrong in multiple times. Gives wrong impression to me.

Personally when it comes to voice acting, I could care less if it's from real person or AI, I have no favorite voice actor anyways. If the speeches are carried properly with correct intonations and consistency, I'd give them a pass.

Do you believe AI voiceover is preferable to no voiceover (including the option to disable it)?

If the voiceover is as bad as in this example, I rather disable it and use my imagination instead.

Should the game notify the player that an AI voiceover is being used?

There's no need.
(+1)

Thank you very much for your response! Again, we decided to include such a voiceover because I know there are visually impaired people who rely on screen narrators to read for them, and I believe that adding different voices would provide a much better experience for them. It is obviously only an option; it is not even enabled by default. It's a visual novel, so you should read it.

This raises a question actually–are screen readers able to read the game’s options? Because if not, sufficiently visually-impaired players may have trouble regardless.

Oh, wow, what an excellent point! I thought it would be a great idea for the game to force you to select a mode before it starts, and if you select the "visually impaired" mode, even the options are voiced. Thank you for your suggestion; there is a lot to consider!

(+1)

Well, using AI to voice characters in a visual novel is certainly a unique approach! I mean, who needs actual human actors when you can just rely on robots to bring your characters to life? I'm sure it's saved you a ton of money on casting fees.

But seriously, I'm not sure how I feel about it. I mean, it's definitely convenient to have AI do all the talking for you, but I can't help but wonder if it lacks a certain human touch. It's like when you try to have a conversation with Siri or Alexa and they just give you robotic responses. It's kind of...creepy.

But hey, if it works for you and your visual novel, then go for it! Just don't be surprised if players feel like they're interacting with a bunch of automated robots instead of actual characters.

Thank you very much for your answer and positive attitude! Sorry that I copied and pasted this answer, but we decided to include such a voiceover because I know there are visually impaired people who rely on screen narrators to read for them, and I believe that adding different voices would provide a much better experience for them. It is obviously only an option; it is not even enabled by default. It's a visual novel, so you should read it.

As for opinion about the human in the voiceover I 100% agree with you.

Thanks :)

Using AI is fine. It's a tool. And over half the people here bitching about "casting saves" are probably using an engine or some other tool that they didn't write to make their games, so screw that hypocrisy.

But in terms of the actual video, where the sampling begins, the voice is alright, but the delivery is flat. It doesn't carry any urgency. You might be able to screw around with some of the sentence structure and punctuation (i.e. "Can you HEAR what I'm saying?" instead of "Can you hearwhat I'm saying?").

(+1)

Thank you for your input! I suppose there are far more computer-assisted things than we can keep track of in our heads. The text and voiceover will undoubtedly be refined during development, but thank you for your suggestion!

I thought it was pretty good! Sure, there were a couple parts where the intonation didn’t match the scene or the pronunciation was off, but I could tell that each character had their own unique voice. If you hadn’t said it was done by an AI I could just as easily assumed that it was done by a human. 

(+1)

This is both terrifying and fascinating. Thank you very much for your input!

Personally, I'd prefer to read the text and would not use the feature. However it does add something to the atmosphere, and may indeed make it possible for visually impaired players to enjoy your game, though it's no substitute for a human VA. Some of the lines have suitable delivery by the AI, but that "omg" at 2:14 is... lacking.

The quality of the generated speech is quite impressive, I must admit. It reminds me a little bit of the voices in some anime where the voice actors weren't always native English speakers and it would show in their performances. I've only used a speech synthesizer for the voice of a spacecraft's computer in my games myself, but I could also see uses for AI speech as placeholders for timing purposes, or while the dialogue hasn't been locked down, until the lines can be recorded by real actors.

(+1)

I literally agree with everything you said. Thank you very much for your thorough response!

(+1)

I personally would have to turn this off in the game, the performances are so wooden, but I'm really surprised how well the voices came out!

I don't think I'll ever be using AI like this, but that's capitalism, I guess! Best of luck with your project. Here's hoping you're able to upgrade to humans in the future.

-E

Hello there! Thank you for your input; I mostly agree with you. I love how you refer to replacing Ai with humans as an upgrade. After all, it appears that the human mind and body are still held in the highest regard.

(+1)

Interesting. I have no idea about quality of this voice... 

(+1)

How do we not know that Carnivore is not a ai, asking for a friend

(+1)

What I think about AI voiceover.  Well. Depends on a lot of things. AI currently gets negative  criticism for ethical reasons, because  for image creation AI is  trained with   images made by    non consenting artists. From what I read about  Replica AI to, they do pay their human AI teachers, so to speak (literally).

So, of course you should not hide the fact that the voices are computer generated. But you need not advertise it either. That is what the about and credits pages are for. Of course you could use it as a marketing ploy. But I for one would be disappointed, if you advetise with "fully voiced" and it turns out, it is not only AI, but most of the text ist unvoiced narrator... I remember being disappointed when some game boasted voice and it turned out  just to be Renpy's text to speach feature that uses the tts from the operating system - of course a big plus for the hearing impaired, if a game actually is made with compability to those features in mind.

Compared to Windows's tts, your example lines were  fantastic. Like the difference between bad pixel art and high res art.  And it is a plus, that the voices for different characters sound different. I remember voice overs from video games 20+ years ago. That was not always done by actors, if you know what I  mean.

So  my verdict in quality is: Text-to-Speach,  Voice Acting by non-actors, The AI stuff, Voice acting by actors, Voice acting by voice actors.

In general, I would prever voice. Lot's of it. But I also frequently hear audio books, so that is a personal preference. For something that might be turned off by the users, it should not be done with the biggest budget. I am sure , that AI voices are a big quality upgrade over text to speach, but it comes with it's own budget and additional work.

What irks me in these type of games is constraints on configuration, that have a negative impact on what is called quality of life (should read quality of gaming ;-). Like not being able to display the text at once, but being reduced to the slow reading speed the    developer hard coded (I am a very quick reader). Also I recently played a game with no back functionality to read the text again, like a text log or an actual back button. Sound has its own qol features, like making it skipable or optional without  disabling the music.