[hts-users:01015] question about HTS lib usage
- Subject: [hts-users:01015] question about HTS lib usage
- From: "Yufeng Hao" <yufenghao@xxxxxxxxx>
- Date: Wed, 28 Nov 2007 11:19:36 +0800
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to:subject:mime-version:content-type; bh=UeeVgu/SycOZxHodFU1isHCX/RGM2GjbkA6fLbIVmEg=; b=pXmRQf7ONtTWrMXQgUJWzoA2ujXEsLLedxCR3JxNsigB3kRUdICKvFcsQobnEVgNMNZGwWpMaBEcFKp44MiCeePfecxQY7F/sBl42P0xcuc23tts8EAbEYQYJKJgqTifXHK4UgIINSv6+8RqhG636mRBUgSzcrKRCV3WEyLgdlI=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=received:message-id:date:from:to:subject:mime-version:content-type; b=sA/B/54tWggl9Y3kedq2pG0+8lxNiUJuooy6gBh8hHSnpfpKBJxxWgLmwn4cU7MCQs7NVETpW0WfDsxnFufhXie89YV5rCfdKp/SmoErMLUT+QP4CFZWL7KOWTYlSz6ceeChVe4eEoL+x5/QQjwzRLKjzf7STxNRDwe+ToLtrL8=
Hi, all.
I have trained HTS-demo sccessful. And I encounter some problems about the trained lib usage.
I looked though "Training.pl", and found that demo sound are
synthesized by HMGENS and SPTK, not with hts_engine builded by festival
and hts toolkit. To synthesizing with hts_engine, I designed the
following experiment:
1. Traing HTS demo with following configurations, and then make:
./configure MGCORDER=24 MGCLSP=0 GAMMA=0 FREQWARP=0.42
2. Write a bash script to synthesize with HTS API v0.9
a). Use the lib result of step 1 as HTS voice model
b). "alice01.lab" in demo/data/labels/gen
script list for example:
./bin/hts_engine -td ./data/model/demo/tree-dur.inf -tf ./data/model/demo/tree-lf0.inf \
-tm
./data/model/demo/tree-mgc.inf -md
./data/model/demo/dur.pdf -mf
./data/model/demo/lf0.pdf \
-mm
./data/model/demo/mgc.pdf
-df ./data/model/demo/lf0.win1 -df
./data/model/demo/lf0.win1 \
-df
./data/model/demo/lf0.win1
-dm ./data/model/demo/mgc.win1 -dm ./data/model/demo/mgc.win1 \
-dm
./data/model/demo/mgc.win1
-od ./data/dur/demo/alice01.dur -of
./data/lf0/demo/alice01.lfo \
-om
./data/mgc/demo/alice01.mgc
-or ./data/raw/demo/alice01.raw -ow
./data/wav/demo/alice01.wav \
-ot ./data/trace/demo/alice01.trace ./data/lab/alice01.lab
Synthesizing ./data/lab/alice01.lab ......
Synthesizing works fine. But the synthesized wave are all slience. I
looked into alice01.trace and found that estimated duration are all
right, but spectrum and f0 are 0 for all frames. I gdb the hts_engine,
but nothing discovered due to the lack of related knowledges. :(
So I wish to know
1. how to use the trained HTS model in festival and isolated hts engine API?
2. does the hts synthesis engine related to the feature-based hts model
? That is, If I train HTS model with MGC-LSP coefficients, should I
synthesize with the same HTS engine API v0.9 ?
Thanks in advance .
Best regards
YF HAO
- Follow-Ups
-
- [hts-users:01018] Re: question about HTS lib usage, Heiga ZEN (Byung Ha CHUN)