how do i make text to speech

[vernal schooner]

i need it in c#, c++, python, c- and i need to also clone unique voice

[hazy wren]

Do you need it in unity though? Is it related to unity in any way?

[vernal schooner]

Unity uses c#

[hazy wren]

OK, but then what do the other languages have to do with it?

[vernal schooner]

I could probably import it using c++ or atleast understand how I can generate realistic

[hazy wren]

Well, anyway, you'll need to use some kind of cloud service, like Google or Open AI tts models. You might be able to run something locally too, but it's probably gonna be resource demanding and low quality. As for the language, if it's a cloud service, you can make requests in C#. Local solutions might require using other languages as well.

[barren yew]

If you do find a library to use locally you want one thats c# or uses c/c++ so you can still interface with it (if cpp you may need to write your own helper functions that are c compatible)

[vernal schooner]

Has to be offline

[vernal schooner]

That's why I'm asking for c, cs, cpp, cm because others are slow. And I need python for another project but it's not required I can just use cpp

[hazy wren]

Well, there are many various libraries. Unity has it's own Ai model inference engine based on onnx. So if you find a model in that format you can use it as is.

Other than that, there's pytorch and stuff. I think Llama cpp also allows tts. You'll need to research. There are just too many options.

[vernal schooner]

I need my own to clone a voice

Not robotic voice

[topaz bronze]

That is going to be an AI model that you train yourself. Once you get a model trained then you can run it locally in Unity withthe Sentis package.

[vernal schooner]

i cant even train text model

with 200gb high quality data

how am i supposed to get proper voicce data

[topaz bronze]

This is getting pretty far outside the Unity world but you can look up XTTS on hugging face

[hazy wren]

Many modern text to speech models allow voice sampling for replicating your voice. Look into the things I mentioned.

[native osprey]

Then you're going to need to lower your standards

[vernal schooner]

clearly it had been done

and not by training it

[native osprey]

Yes, using existing text-to-speech text banks that you've decided aren't good enough for you

[vernal schooner]

they dont clone voice

and they dont use proper language

like c++ or c#

[native osprey]

Right. And "Cloning a voice" isn't something you can just do

It usually involves training a large AI model on a bunch of clips

which you've also refused

So, you need to either do that, or lower your standards

[vernal schooner]

it had been done without using large ai model

nobody uses it

except for microsoft

[native osprey]

Not really. Text to speech is either using recorded voice banks, or AI

Both of which you've refused

[vernal schooner]

what do u mean by voice banks

[native osprey]

Lots of recorded sounds, either entire words or phonemes, spliced together to make the sentence you've typed in

A.k.a, that thing you said you didn't want to do

Because you want to clone a voice

[vernal schooner]

how will that allow different tones and avoid extra letters like h

[native osprey]

By having a large voice bank

[hazy wren]

You don't have to train a model from scratch. You can fine tune one on a smaller set of data. That should be more achievable with limited hardware and probably provide better results than existing voice cloning capabilities(although I've seen some voice cloning/sampling examples that sound quite impressive).

As for the programming languages, I don't see how that's a problem. You can have an interop with virtually any language. At least if you're targeting pc/mac platforms.

[oblique vine]

I wonder if its even possible in the first place to clone your voice exactly, without training an AI
Prior to a lot of the recent AI TTS examples, the voices often sounded robotic and clearly synthetic

[topaz bronze]

Without AI, you can only reproduce the phonemes you have recorded. At best you get the star trek computer voice. Or Siri I guess.

[vapid skiff]

what is your big picture goal?