Synthesising directly from a phone sequence rather than text

[Etienne D]

Hi,

I’m attempting to perform synthesis from a sequence of phones. I have tried to use the SayPhones method but I could not find enough documentation: the following (and other phone sequences as well), taken from a random paper on the web, only yields an EST error “item is null so has no stress feature”.

festival> (SayPhones ‘(# n o t w @@r k i ng #))

I then stored a sequence of phones into a variable like so (using Utterance Phones):

festival> (set! someutt (Utterance Phones (# h @ l ou #)))

and tried to construct the utterance step by step in a similar fashion to what we had done in the 1st Speech processing assignment, but failed again.

Could you provide some explanation as to how to go about synthesising from a phone sequence, and optionally, how to fine-tune parameters such as length and stress?
Thank you,
Étienne

edit: fixed the link

[Simon]

I’m not sure of the solution to this. Let’s talk in person – is Festival the best framework for you, or should we consider a DNN system?

[Simon]

SayPhones is probably only going to work for a diphone voice, not a Multisyn unit selection voice. Try loading a diphone voice and see if that works. You are going to get monotonic F0 though, I think.

[Author]

Posts