Probably work! I think there's some difference in the trig, and functions like ceil and floor. I'm going through each function printing values as I go.
I'd like for the data to be the same, although I know double precision is going to be a bit different than single, but I shouldn't be able to get a pitch value of 31 on one machine, and 18 on another :-) When I find it, I'm going to kick myself, I know... On Tue, Dec 15, 2015 at 4:20 PM, Ricardo Andere de Mello <[email protected]> wrote: > use bigdecimal? > > 2015-12-15 18:51 GMT-02:00 Steve <[email protected]>: >> >> I've decided that the hts1.raw file is indestructible and always sounds >> good :-) >> >> I've been trying to find the bug in my java codec2 version, but even >> with outputting completely different data in the .c2 file, the hts1 >> voice still sounds good! >> >> Course in real-time, my voice sounds like crap, but I've decided that >> getting reliable output from the test files should probably precede >> real-time testing. >> >> An example .c2 error: >> >> original: c066 17e0 >> LSP[110000][00 0110][0110 00] Energy[010] Pitch[111 11] Voicing[1] >> Spare[00000] >> >> mine: c017 1480 >> LSP[110000][00 0001][0110 00] Energy[010] Pitch[100 10] Voicing[0] >> Spare[00000] >> >> I hate floating point... >> >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> Freetel-codec2 mailing list >> [email protected] >> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > > > ------------------------------------------------------------------------------ > > _______________________________________________ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > ------------------------------------------------------------------------------ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
