Hello all,
I am a student currently working at UPC (Barcelona, Spain) trying to
build some free catalan voices.
I am trying to setup HTS-demo ADAPT scripts to use them with our
speakers and I have a question:
In ./configure --help :
[...]
Some influential environment variables:
[...]
SPEAKER speaker name (default=slt)
DATASET dataset (default=cmu_us_arctic)
TRAINSPKR speakers for training (default='awb bdl clb jmk rms')
ADAPTSPKR speakers for adaptation (default=slt)
[...]
My question is: What is SPEAKER for?
As far as I know there are some speakers 'TRAINSPKR' to train the
"generic voice" and then this "generic voice" is adapted to other
speakers 'ADAPTSPKR'.
Maybe I am wrong but the only place I have seen SPEAKER used is in
scripts/Config.pm at the beginning: $spkr. I think that $spkr is only
used at "converting mmfs to the hts_engine file format" in
scripts/Training.pl.
I am not sure, but I guess that this step was copied from
HTS-demo_CMU-ARCTIC-SLT and it has to be adapted... Am I right?
I can try to adapt it somehow, but I am not familiar with the code so
maybe my solution is not going to be the best. Can anyone help,
please?
On a different topic, I am adapting the extension "Multi-process
supporting patch for HTS-demo_CMU-ARCTIC-SLT" to adaptation scripts.
If I succeed I will submit the patch both here and to Xuchen Yao (the
original creator of the extension).
Thanks for your time, and forgive me if the question was a bit dumb :-)
Sergio Oller
Universitat Politècnica de Catalunya