Chinese Synthetic speech pacing is very fast

synthetic speech pacing is very fast B11_2.zip

for chinese language , anyone have same test result?

How about your training data speed?

my training data speed is normal. when i use the train data which is generated by forced aliment to genarate wave , its speed is normal.

when i use the test data which is genarates by fronted to generate wave, its speed is fast.

i find tha train data and test data with same text have different lab.

There’s something wrong with downloaded train data lab file. Use front end to generate new lab file.

Cannot hear anything.

@v-yunbin a Chinese Front tool Chinese Front toolcan generate lab file like this:

0 0 a4^k-uai4+w=uen2@/A:4-4^2@/B:7+2@2^3^2+9#2-9-/C:n_n^u#0+1+0&/D:xx=10!xx@1-1&/E:xx|10-xx@xx#1&xx!1-1#/F:xx^10=17_1-1!
0 0 k^uai4-w+uen2=zh@/A:4-2^1@/B:8+1@3^2^3+8#3-8-/C:n_n^u#0+1+0&/D:xx=10!xx@1-1&/E:xx|10-xx@xx#1&xx!1-1#/F:xx^10=17_1-1!

namely ,there is no duration information before phone label.Could this kind of labe file be feed into merlin to synthsis voice? Or, what step further operation should i do to fill duration ahead of phone information? The real lab file may look like this

26500000 27500000 a4^k-uai4+w=uen2@/A:4-4^2@/B:7+2@2^3^2+9#2-9-/C:n_n^u#0+1+0&/D:xx=10!xx@1-1&/E:xx|10-xx@xx#1&xx!1-1#/F:xx^10=17_1-1!
27500000 28300000 k^uai4-w+uen2=zh@/A:4-2^1@/B:8+1@3^2^3+8#3-8-/C:n_n^u#0+1+0&/D:xx=10!xx@1-1&/E:xx|10-xx@xx#1&xx!1-1#/F:xx^10=17_1-1!