Thank you for you suggestion. And also I have one more question, after I follow the step, I found that the final output quality is but. It mentions that Getting good pitchmarks is important to the quality of the synthesis, and I review the information in this web site: http://festvox.org/bsv/x862.html, and compare my .pm file, the graph is very different, see here http://i43.tinypic.com/28i450h.png, there are many red lines, and I open my .pm file, the content is sth like that:
EST_File Track DataType ascii NumFrames 466 NumChannels 0 NumAuxChanels 0 BreaksPresent true file_type 14 EST_Header_End 0.010042 1 0.020083 1 0.030125 1 0.040188 1 0.050229 1 0.060271 1 ...... ...... Is my .ps file correct? Since there are too many pitch mark..... if not correct, how can I fix it? Actually I have edit the script bin/make_pm_wave, but the output is still bad... Thanks and regards, Vincent 2012/2/17 Nickolay V. Shmyrev <[email protected]> > On 17/02/2012 at 01:45 +0800, Siu Meng Lei wrote: > > I am following the step of this web site to new the new language > > voice : > > http://festvox.org/bsv/c3456.html > > Instead of using this command to record voice: > > bin/prompt_them etc/jadiph.list > > I use other software to record my voices independently and get the > > WAV > > files. > > After that I put those WAV files back to Festival and run the > > following > > command: bin/make_labs prompt-wav/*.wav > > however, it shows the following message: > > ja_0001 LABELING > > RIFF file: unknown sample format > > How can I fix the problem? > > (my WAV file format is Uncompressed 32-bit IEEE float audio, Sample > > rate > > 48000 Hz) > > If I change the format, does it affect the quality of result? > > Thanks and regards, > > You need to change format to 16khz 16 bit mono signed little-endian. > > You can do it with sox. > > It will not affect the quality > >
_______________________________________________ Festlang-talk mailing list [email protected] https://lists.berlios.de/mailman/listinfo/festlang-talk
