Faceposer Phonemes

[Tynan]

Faceposer is now refusing to properly load the phonemes. It used to work fine, then it worked intermittently, now it never works. All the phonemes come out as blank expressions, which means that my character are saying their lines, but their lips don't move because all the morph targets are the same blank look. Does anyone have a solution please? I've been working on this for some time.

 

[Sodabus]

I'm not exactly sure what the problem is.

Are you talking about the phoneme extractor? Are you getting the infamous "An error occured during extraction" and "...full recognition" errors? Do <sil> phonemes pop up everywhere?

 

[Tynan]

The phonemes are all there, it's just that they're all blank expressions. As in, when I remake my BMPs all the phonemes are just blank looks, plus the models don't move their mouths when speaking. Custom made expressions work fine.

 

[Sodabus]

Check out this link; I think this person had the same problem as you.

Problem with phonemes.txt

 

[drake]

Unrecognized phoneme

I'm not exactly sure what the problem is.

Are you talking about the phoneme extractor? Are you getting the infamous "An error occured during extraction" and "...full recognition" errors? Do <sil> phonemes pop up everywhere?

Sorry to bump this ancient thread, but I've run into the 'full recognition' problem, and am having trouble finding a solution to it. The symptoms are the same as Sodabus described: Every time I load a wav file and try to extract the phoneme data, I get a ton of 'Unrecognized Phoneme' errors in Output, and the result is '<sil>' phonemes scattered all over the waveform. (I've never gotten anything other than <sil>, not even a 'wrong' phoneme.)

I get this result regardless of which WAV I load. When trying the SDK's sample WAV, I notice it automatically extracts the words and phonemes perfectly, so I assume these are embedded in the WAV file. However, if I try to re-extract the phoneme data, I'm rewarded once again with <sil> all over the place.

My WAVs are correctly placed in steamapps\myuser\half-life 2\hl2\sound\. I have 2 custom wavs, one by myself and one by a professional. They're all in fine quality, and I've adjusted the volume to avoid the [! Too loud] error. I'm also using sdk_al_showonroad.wav when testing.

My SAPI is 'Microsoft Speech SDK' version 5.1. I have it configured for medium accuracy on 'Microsoft English Recognizer v5.1'. My SAPI profile is untrained, but it does recognize my speech to some degree. (I've tested Dictation in MS Word, and the words I get sound similar to what I say into the microphone.)

My microphone is of good quality.

Faceposer CC Language is 'English'

Faceposer Expressions are working fine (so this is NOT the problem Tynan had; it's the one Sodabus hinted at.) The model will move its lips when I play the sample WAV for the first time or move my mouse over valid phonemes, so it has to be a recognition problem.

I hope someone can help me with this; it's a 'This should be working...' situation and I'm a bit stuck. Thanks in advance.

 

[drake]

Well someone might get this problem again in another 2 years, so here we go:

It turns out the Phoneme Editor is not configured to work with non-U.S. versions of MS Windows, which causes conflict between Faceposer and the SAPI, (even if you change regional settings etc.)

By a stroke of luck, 'Qyon' at the ChatBear forums has written this small program, which patches the Windows Registry so Faceposer will communicate properly with other languages. You may have to disable virus protection for it to run.

If this still doesn't work, your last resort is to buy a U.S. version of Windows. Still, thanks to Qyon for this one!