Voice recognition in godot

hjiul

Hello,

I have a client that want to work in a sort of educationnal game to learn Hebrew.

It will involve to learn to pronounce letters correctly.

As i love working with godot, I wonder if there is some ways to make that possible.

Is this one the way to go : https://github.com/menip/godot_speech_to_text

Have the voice converted to text and then compare the character to see if its correct.

Is there a more simple approach ? I know i could just record the voice and send it to an api. But the cost of using such api is pretty big so I dont think its a valid option.

Another approach would be to compare audio files.

I could have an audio file of each letter or word and then compare it to the recorded input.

I guess its pretty simple to check if they are eaxctly the same. But is their a way to detect if they are "similar enough" ? And if yes, how to set the margin of error we allow ?

Thanks a lot for your help :)

cybereality

It's a really hard problem, even more so if it's not a popular language like English or Spanish. There have been decades of research, and it still doesn't work all that great. I don't want to discourage you, but it's a huge undertaking and not an easy thing.

hjiul

Totally what i thought :)

SO using an api is the only realistic approach, right ?

cybereality

Realistically, yes. Using an API would be the easiest thing. Also, the speech to text plugin is one option if you can't afford the API (I know they are really expensive from when I looked into it). But Sphinx doesn't support Hebrew, which means you would need to create the language model yourself, and that could be time consuming and costly.