Using AI for Speech Packs?

[+allenjb42 4,234]

In the modern world of deepfakes and AI generated speech, would it be possible to leverage some of the tech out there to create additional speechpacks for SF2?

I'm thinking packs for various international accents of English, as well as other languages.

Are there any programs out there that you could feed the stock speech text (is there a file with that somewhere?) into and then get the required speech files in the language, gender etc. that you wanted?

Cheap programs would be good. Free would be better.

Any thoughts?

 

[mono27 85]

That would be a great idea. We already have many languages in speech packs here in CA. But if we can have more it would be even better!

[alanthegreat 1]

I would like to say, that would be great. To have the sound be similar and not jarringly different that happened in StrikeFighters2 NorthAtlantic. Also having adding  new voice and contact from wingman's would be great. Please do this maybe I can help.

[+Menrva 4,200]

I've already thought about this for the ODS 30AE mod, but you'd need to spend money for those AI services that generate audio speech files, I haven't found viable freeware sources, and you'd also need a lot of patience and free time as a single complete speech pack requires more than 1,400 audio files. Do it for several languages, and you'd need to multiply that amount for each language/accent you intend to have. Those services cost money and the ones I had seen do not allow to automatize the creation of several audio files in one go, so they also cost in terms of time. I'll probably consider it at a later time, but that would mean spending my own money and what little free time I get over something that only few Strike Fighters fans would appreciate all for free.

EDIT: That being said, it surely is the easiest way we currently have to make a more professional speech pack, but not as easy as I'd like.

[+daddyairplanes 10,233]

my two cents,  dont use AI

i dont know how it is with audio files at this time. but pure AI images (type in a description and an algorithym produces an image) are currently kinda junky. between the weird FB images posted for likes to a proliferation of images ran through a filter to "clean up the image" which results in something that looks like a water color of the original at best.

but the more that it is improved, the more likely that companies will go to it for producing artwork.

if theres one thing that should not be lost to automation, its human creativity.

time is always a factor in everything, esp. modding. but we have a global community here at CA. lets use any of our non English speaking partners who have the time and are willing to do so make up speech packs.

besides, that way we'd also get teh benefit of their local accent on the pack instead of a computer generated speech. it might not sound too different to us native English speakers, but would be very noticable to the Arabic, Spanish, Chinese or whatever other language speakers you are looking for.

Edited September 16 by daddyairplanes

[+EricJ 4,241]

16 minutes ago, daddyairplanes said:

my two cents,  dont use AI

i dont know how it is with audio files at this time. but pure AI images (type in a description and an algorithym produces an image) are currently kinda junky. between the weird FB images posted for likes to a proliferation of images ran through a filter to "clean up the image" which results in something that looks like a water color of the original at best.

but the more that it is improved, the more likely that companies will go to it for producing artwork.

if theres one thing that should not be lost to automation, its human creativity.

time is always a factor in everything, esp. modding. but we have a global community here at CA. lets use any of our non English speaking partners who have the time and are willing to do so make up speech packs.

besides, that way we'd also get teh benefit of their local accent on the pack instead of a computer generated speech. it might not sound too different to us native English speakers, but would be very noticable to the Arabic, Spanish, Chinese or whatever other language speakers you are looking for.

I agree, I don't use AI in my books at all.

[+Menrva 4,200]

2 hours ago, daddyairplanes said:

if theres one thing that should not be lost to automation, its human creativity.

I couldn't agree more. It's unfortunate that proper artists are losing jobs over sub-par, soulless AI creations. AI should be a very last resort when it comes to making art, it shouldn't replace the artistry of people from all over the world. AI generated images do not give me the same vibes. I have great admiration for modders who create exquisite models and textures, throughout the years they encouraged me to learn more and do something of my own creation. AI generated content does not inspire me to grow and learn more.

[simonmiller416 1,830]

I'm trying to do something like this,in this case I'm open to AI because the Strike Fighters fan base is very small, that's a fact.

In my language area, there are very few players, and even fewer people who can voice the characters. So I'll let AI do most of the work.

Edited September 17 by simonmiller416

[Stratos 3,188]

On 17/9/2024 at 5:24 AM, simonmiller416 said:

I'm trying to do something like this,in this case I'm open to AI because the Strike Fighters fan base is very small, that's a fact.

In my language area, there are very few players, and even fewer people who can voice the characters. So I'll let AI do most of the work.

I'm curious now, are you Chinese?

[+yakarov79 11,082]

Well it is your choice how you will proceed, but one thing for sure when using so called AI* - you will not gain experience. At least not the same way as making it on your own, or maybe with some new people from another countries, that might become your friends. *(I still refuse to accept it as AI, yes it is very sophisticated program, but very far from intelligence, artificial intelligence). 

Fortunately era of internet give us tremendous opportunities, in meeting, contacting and cooperating with people from another countries almost instantly. Just need to reach to the right person.  I remember some 30 years ago (the days when Internet for me was some mysterious entity) when I was researching some info about CBI theater Mustangs for my yet another scale model Mustang. Few curious questions reach right persons, on right forums and at the end with cooperative research we unearthed many interesting facts about some squadrons, airframes and servicemen in CBI. And thankfully i gained some experience, but what is most important I gained many dear friends, some I never met in person yet still we reach out sometimes just to say simple Hi and "whats new in aeroarchaeology"  because there is always something new to discover. .. And same on CA, working on some projects with guys here gave me unforgettable memories - salute to you all- you know who you are,.

As for the AI, fun thing is, and very sad that nowadays humans now are so impressed with art created by so called AI. Well, thing is I do not give a damn what software can create - it will be still based on styles created by humans. Yes, the same way as artists influence another generations of artist. But it is us humans who should be creative and most importantly - have time to be creative, have time to explore. While so called AI should do the simple basic things so we humans could find more time for that art and to be creative. Sad thing is that "AI developers" see things different way. This is not the way... skynet way , no way. Anyway.

 

 

[PeacePuma 569]

I am afraid that this time I disagree not entirely but with the context that has been given to what the guys are proposing (@allenjb42 and @simonmiller416) , I will explain my reasons; The famous AI let's see it for what it really is, one more simple tool that we can resort to to reduce times or be precise in areas where we cannot be, but in no case will it replace anyone's talent, to see this as a detriment is to make oneself replaceable, like you guys too I hate auto tune, Canva, web design templates, and the invasion of deepfakes on social media, but this "is not the fault of the pig but of the one who feeds it" we are the users who make excessive use of these solutions, I started using graphic applications at the dawn of this century and many things have changed with filters and automations and I will not fall into purism because another guy saves time.  "The AI art" does not supplant anything, art itself is an expression that evokes sensations in the spectator, only you know if you are satiated with a hotdog or a crème brulee, technique and talent is part of the pack of the artist that I love, ai is the one that like the boy who makes a living on the street painting similar landscapes with spray and stencils, I will not judge him as an artist but like AI I will say that he has technique and nothing else.

Let's go back to the original topic, let's be honest, it has nothing to do to depend on a tool, with the fact of making art in a language pack where we know it is very complex to get someone to do the voices in different languages, it's like judging you for using a translator... hey this same forum has lowered the enthusiasm to those who propose to buy the source code of the game for underestimating costs and now we are going to discourage people who do not have money to pay speakers to do an eternal work.. have you seen the amount of audios the game has? let's consider that repeating “Red Crown” or “bandits at 3” in different accents or languages will not give me a great special talent (maybe learn some language) but I would hardly learn the accent of a language that I don't speak natively... I prefer to leave that to an AI that will not do it brilliantly but much better than oneself... In my opinion, if you can do it, do it cowboy, it's better than waiting for nothing... and what I can contribute is that the ChatGPT mobile app has an advanced voice mode (paid) that imitates the most characteristic accents of the languages, if you must solve zone limitations that surely if you are from china you must know how to solve (sorry Simon, I really don't know if you are from china) XD 

Greetings!

Edited October 4 by PeacePuma
the ia also makes mistakes