TEXT->REALISED_DURATIONS works, REALISED_DURATIONS->AUDIO doesn't

[kjyv]

As with other combinations, some of the generated data seem to not be parsed properly.
While it is possible to synthesize REALISED_DURATIONS from TEXT or EMOTIONML,
generating audio from these is not possible (server gets stuck with 100% cpu).
An example input is

0.080000006 125 h
0.14000002 125 a
0.19500002 125 l
0.30500004 125 o:
0.64500004 125 _

It would be very helpful to have this fine grained control over the output.

[alishah-ahmed]

Could you please provide us with the following details if possible:

• Stacktrace
• Java Version
• Machine details that you are using along with OS details.

Thanks.

[kjyv]

As it "gets stuck", there is no stacktrace, as no error is thrown. At least I couldn't see any on both terminal windows (server or client).

$ java -version
java version "1.7.0_72"
Java(TM) SE Runtime Environment (build 1.7.0_72-b14)
Java HotSpot(TM) 64-Bit Server VM (build 24.72-b04, mixed mode)

Running on OSX 10.10.1.

[psibre]

Could you please provide the details of your request? Which locale/voice are you using?

[kjyv]

Language de or en, and it seems the voice doesn't matter, bits1-hsmm, bits3, bits3-hsmm, dfki-pavoque-neutral-hsmm all produce this behavior.

[psibre]

I believe that the source of this issue is that REALISED_DURATIONS and REALISED_ACOUSTPARAMS were intended only as output types, but when they were originally introduced (in bd582c8 and eb8807b, respectively), their isInputType parameter was set to true.