Another thing that is confusing is that when I say for example festival> (SayPhones '(l d))
I get a badly distorted d sound. But when I play the file us_1100 which the diphone index has mapped to l-d it sounds fine and the label looks fine as well. If I replace the mapping in the index for l-d with l-dh it sounds fine, no distortion. So if the original recording is not distorted where is the distortion coming from? On Mon, Jun 20, 2011 at 9:55 AM, Heather Dewey-Hagborg < [email protected]> wrote: > Thanks for the advice, I will check the labeling more thoroughly. spot > checks looked pretty good but I didn't review every recording. > One issue I have noticed is that the script which makes the index doesn't > properly add the diphones like p_-_l timings. They just appear as 0 0 0 even > though the lab file is correct. Anyone know a fix for this short of adding > all the hyphenated diphone timings by hand? > > > On Mon, Jun 20, 2011 at 2:32 AM, Jeremy Salwen <[email protected]>wrote: > >> Have you checked the labellings of your prompts? >> >> To do so, copy the contents of your wav folder and the contents of your >> lab folder into the same directory (or setup links to make it seem that >> way). Once you've done that, open up the wav files with waveurfer, and >> choose the "transcription" view for all of them. Now you can go through one >> by one and check if the labellings are right. Options are: re-record the >> ones with bad labellings (remember to run bin/make_lab again before checking >> the labels again, I made this mistake once, and kept re-recording and >> thinking that the autolabeller sucked. Also, to save time, you can run >> bin/make_labs prompt-wav/test001.wav to just relabel test001.wav, instead of >> doing it to all the recordings, which can be time-consuming.), or >> hand-correcting the labels. You can literally just drag the labels from >> within wavesurfer (remember to copy your changes back to the lab/ >> directory). >> >> Once you've got all the labels as perfect as you care to have them, just >> repeat all the steps after "bin/make_labs prompt-wav/*.wav" and you should >> get the voice built with proper labeling. >> >> Jeremy >> On Sun, Jun 19, 2011 at 7:42 AM, Heather Dewey-Hagborg < >> [email protected]> wrote: >> >>> So I have successfully gone through all the steps here: >>> http://festvox.org/bsv/c3619.html >>> to build a new US voice and it actually speaks! >>> >>> The problem is it sounds almost indecipherable... I think many of the >>> phones are bad - either the recording is bad or the alignment is bad. Some >>> phones sound just like noise or hiss or snaps pops etc. some phones sound >>> good. >>> I have attached a wav file of the voice saying "once upon a time in a >>> land far far away hello world" >>> in case anyone has troubleshooting tips. Is there a way for example to >>> have festival say what phones it is using for text input? Or to test every >>> phone in the voice database? Or to trace text input to the original recorded >>> file in some other way? any advice appreciated... >>> >>> thanks, >>> Heather >>> >>> _______________________________________________ >>> Festlang-talk mailing list >>> [email protected] >>> https://lists.berlios.de/mailman/listinfo/festlang-talk >>> >>> >> > > > -- > Heather Dewey-Hagborg > > www.deweyhagborg.com > 518-598-3775 > -- Heather Dewey-Hagborg www.deweyhagborg.com 518-598-3775
_______________________________________________ Festlang-talk mailing list [email protected] https://lists.berlios.de/mailman/listinfo/festlang-talk
