I’m new in Unity 3D. I’m makeing a game like “Talking Tom Cat”. Actually I’m gonna record user’s voice and make it like a specified human voice (that I have his voice !).
I can record user’s voice, but I don’t know how to make it like another human voice ??!! I want it for android devices.
Would you please help me how to “SIMULATE SOMEONE’S VOICE” ?!
The effects section of Unity has the Pitch shifter, which has the option to shift the pitch without actually changing the timing. This is what these Talking Cat like games use.
Can you give an example? I’m quite experienced at audio, but I don’t think that’s possible, if I’m thinking of what you’re thinking.
You are facing two problems:
A voice is a very complex composition of timbres. That’s why robot voices still don’t sound natural. So that’ll sound weird, even if you have the best of tech AND voice sets to mimic that.
If you want a synthesised voice, you will ALSO need something to interpret what has been said, which is also terribly complex. That is something Google does when you say what you want to search, and it converts it to text (which can be used in your example for the synthesised voice)
I couldn’t find an example on YouTube, but I don’t think these repeat-what-I’m-saying-apps are doing.
OK but did you see the video? He already does a basic impression of Morgan Freeman. I don’t mean to be rude, but forget about the idea. It’s not possible.
AliRezaBeytari, getting it to sound even remotely like your voice is still incredibly complicated. There are some nice EQ and pitch tricks to go from one specific voice to another quite similar one, but it has to be carefully configured specifically for the input voice. Additionally, the input voice has to have similar characteristics as the target voice, and the person must specifically attempt to speak in a similar manner. The worse the match is between the input and target, the harder the process is going to be. With a poor match, even an audio professional with years of experience may be unable to pull it off even with the craziest formant modulation tricks, and this is still a scenario where a person is working with a specific set of recordings. You, however, need an algorithm to figure this out on its own behalf, regardless of input voice, in realtime. My best recommendation is that you practice to work with audio. It won’t solve the problem, but maybe it’ll help you come up with something else that works for the game? As a nice bonus, you’ll learn a lot of useful skills.