[hts-users:03248] HTS-demo_CMU-ARCTIC-ADAPT: new target speaker / hts_engine files not correctly generated
- Subject: [hts-users:03248] HTS-demo_CMU-ARCTIC-ADAPT: new target speaker / hts_engine files not correctly generated
- From: Pierre-Edouard Honnet <cerbrain@xxxxxxxxx>
- Date: Tue, 17 Apr 2012 12:25:39 +0100
- Delivered-to: hts-users@xxxxxxxxxxxxxxx
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:from:date:message-id:subject:to:content-type; bh=DH2QjMRiOZXOmuB1JiI5b92MVcv5pr3tDCSRAvGu3A4=; b=L/mZhBEZWXkdDMXcL1OgPAPcvMX3G54AIB4co+Th3AF7vNIaF2Pq3TpLe4M/QkYoRh vbMXWHV7Qraxhai3og+quXgQwiaRuClSSoXaAtiu9VDkdpDs2A7DVKcqjIyZLeNo4x2B 0r/JyYWYc8KmD5WBU3fAQFK7E2rmc9MxOfFsO1wht3B8IkLpZtiLjDSDmK8gJInQrgNB VqZ788FTyyzf+/cM9mBfN3mG0BzI/hNFQ9Ix/fy0PY4QTVaBKQM5MgVUI1lMhHhhUEyf YnOtDB35fx2LwVepNs+/NvL7+qFSKCzPuHI5PtvMIHu+Xa9A/siqapGDDAK6mGwcHlbh qoWg==
Hi all,
I have 2 questions about the HTS demo for speaker adaptation:
- I managed to change the target speaker for the adaptation choosing one of the 6 speakers from CMU Arctic database, however I would like to adapt the average model to another speaker. The problem is that I don't have the same sentences in my own database, and it seems that I can't use my labels for this script. Here is the error I get:
Processing Data: CLCS_SCL_AF430013_04.cmp; Label CLCS_SCL_AF430013_04.lab
ERROR [+7321] CreateInsts: Unknown label x^x-pau+ay=w@x_x/A:0_0_0/B:x-x-x@x-x&x-x#x-x$x-x!x-x;x-x|x/C:1+1+1/D:0_0/E:x+x@x+x&x+x#x+x/F:content_1/G:0_0/H:x=x@1=3|0/I:5=5/J:14+13-3
FATAL ERROR - Terminating program /usr/local/HTS-2.2beta/bin/HERest
So I wonder if my problem is that I have some unseen phone association (but when I looked at full.list it seems that they are in the list) and how I could solve this problem or if it is not possible to adapt the voice with other labels than the ones of CMU database ? (in the latter case, is there simple solutions to adapt with non parallel corpus ?)
- At the step when mmfs should be converted to the hts_engine file format, it seems that there is a problem although I didn't change the script, and in the last step when synthesizing the waveforms using hts_engine I got a problem, even if I try to use different versions of hts_engine, I guess that the script was written for a previous version of hts_engine (using different header) and that this part has not been adapted. Here is the error using hts_engine 1.06:
===============================================================================
Start synthesizing waveforms using hts_engine at Sun Apr 1 07:19:54 IST 2012
===============================================================================
Synthesizing a speech waveform from /home/fred/peh/HTS-demo_CMU-ARCTIC-ADAPT/data/labels/gen/cmu_us_arctic_slt_alice01.lab using hts_engine...
Error: HTS_Model_load_pdf: Failed to load header of pdfs.
I get the same error with hts_engine 1.05, and this one with hts_engine 1.00
===============================================================================
Start synthesizing waveforms using hts_engine at Mon Apr 2 12:04:21 IST 2012
===============================================================================
Synthesizing a speech waveform from /home/fred/peh/HTS-demo_CMU-ARCTIC-ADAPT/data/labels/gen/cmu_us_arctic_slt_alice01.lab using hts_engine...
Error: HTS_Model_load_pdf: MSD weight should be within 0.0 to 1.0.
Maybe someone has a solution to this problem?
Regards,
Pierre-Edouard Honnet
- Follow-Ups
-
- [hts-users:03249] Re: HTS-demo_CMU-ARCTIC-ADAPT: new target speaker / hts_engine files not correctly generated, Keiichiro Oura