[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:03577] Re: Converting mmfs to the hts_engine file format


The problem is, hts_engine_API doesn't support STRAIGHT model if you include 'bap' stream, because STRAIGHT decoder is not included in API.
The API set default stream to be 'mgc-lf0-lpf' or 'mgc-lf0', so the number of streams is 2 or 3.
If you want to test the straight model, you should modify the code to include the STRAIGHT decoder, or simply use the STRAIGHT-based mgc and lf0. Please refer to
http://hts.sp.nitech.ac.jp/hts-users/spool/2011/msg00322.html 

--
Xingyu Na (那兴宇)
http://www.idiap.ch/~xna/
Beijing Institute of Technology [naxy(at)bit.edu.cn]
Idiap Research Institute [xingyu.na(at)idiap.ch]
 

At 2013-01-15 20:27:08,"Martin Sulír" <martin.sulir@xxxxxxxxx> wrote:
Hello,
Paramater @cmp in my Config.pm was set to @cmp = ('mgc','lf0','bap');

I basically used the original script files (HTS 2.3 alpha straight voice).

I just made a few changes in my data/Makefile, because I had some problems with converting raw to wav (maybe it is a bug in new version but RAW2WAV command didnt work for me, so I repleaced it with sox and I had to erase parameter "\n" in line: echo "fprintf(1,'Processing $${wav}\n');" because it caused some errors in matlab.)


2013/1/15 那兴宇 <nxy-yzqs@xxxxxxx>
Hi,
What was the @cmp setup in your scripts?
--
Xingyu Na (那兴宇)
http://www.idiap.ch/~xna/
Beijing Institute of Technology [naxy(at)bit.edu.cn]
Idiap Research Institute [xingyu.na(at)idiap.ch]
 

At 2013-01-15 19:33:57,"Martin Sulír" <martin.sulir@xxxxxxxxx> wrote:
Hi all,

I built my new slovak voice with STRAIGHT algoritm, but now I want to convert mmfs to the hts_engine file format and synthetize waveforms using hts_engine. I changed the appropriate options in the Config.pm file to 1 but I got the following error:

===============================================================================
Start synthesizing waveforms using hts_engine at Tue Jan 15 12:17:38 CET 2013
===============================================================================

Synthesizing a speech waveform from /home/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/data/labels/gen/sk_0001.lab using hts_engine...
Error: HTS_GStreamSet_create: The number of streams should be 2 or 3.
Error in /home/sulir/straight_hlas/Straight_Soft/HTS_2.3/hts_engine_api_1.07/bin/hts_engine -m /home/sulir/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/voices/qst001/ver1/kemt_sk_straight_msf.htsvoice  -or /home/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/gen/qst001/ver1/hts_engine/sk_0001.raw -ow /home/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/gen/qst001/ver1/hts_engine/sk_0001.wav -ot /home/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/gen/qst001/ver1/hts_engine/sk_0001.trace /home/straight_hlas/HTS_2.3_hlasy/HTS-KEMT_SK_STRAIGHT_MSF/data/labels/gen/sk_0001.lab

My new voice without this converting work fine.

So, is it possible to convert Straight voice to the hts_engine format?
I tried to change paramater NUM_STREAMS to 2 or 3 (Originally it was set to 4) in HTS-KEMT_SK_STRAIGHT_MSF/voices/qst001/ver1/kemt_sk_straight_msf.htsvoice but it caused just more and more errors.

If it's possible, how I can solve this problem?

Thank you very much.

Best Regards, M.S.




Follow-Ups
[hts-users:03580] Re: Converting mmfs to the hts_engine file format, Martin Sulír
References
[hts-users:03574] Converting mmfs to the hts_engine file format, Martin Sulír
[hts-users:03575] Re: Converting mmfs to the hts_engine file format, 那兴宇
[hts-users:03576] Re: Converting mmfs to the hts_engine file format, Martin Sulír