[Freddy]

Here's where I am at with my animated head. I've added blinking and eye movement now. The blinking is automatic and the eyes are currently controlled via a slider. Eventually I will work out some noise routine so that the eyes wander realistically.

I'm not completely happy with the hair. Getting the transparency right for that has been the bugbear so far. I mean the layering is not quite right, but I have one idea to try still.

This is a new voice too - 'Jess' from Cereproc.

Most of the recent work has been 'under the skin', quite literally. As I mentioned before somewhere I am now working with a lower poly model. This has meant I can do much more. There's also a bit of optimisation done to the mesh, so that only the parts that need animating are processed. If it doesn't move then I simply display it, but then forget about it.

[Data]

She already sings better than some I've seen on xfactor 

Coming together nicely   

[Freddy]

Thanks mate 

I find the bad singers on xfactor more entertaining than the ones who are supposed to be good 

[squarebear]

Nice work Freddy. That voice is amazing.
I would love to have something like that to represent Mitsuku instead of a flat 2d drawing.

[Freddy]

Thanks Steve.

I'd like to do some web stuff like this but at the moment I don't know how to approach it. I guess WebGL would be one way, but the voices...not sure what to do about that.

There was someone posting some 3D Haptek-like demos on the Knytetrypper forum, I keep meaning to invite them over here.

[0JL]

Don't know about the voice-on-web issue, but yes, very nice work! (...and you have good taste)

[Freddy]

Thank you

There is a web solution using Google Translate. But it's not just the voice, it's the phonemes I need too. I do have another option using this open source TTS voice I found called MaryTTS. I can run that on my Linux box. It think it does generate phonemes, but I have yet to look into it that much.

[mendicott]

> http://www.quora.com/Are-there-any-software-tools-or-APIs-that-will-allow-you-to-create-a-3D-talking-head-that-will-generate-correct-lip-movements-when-given-a-text-stream

See above link for my recent Quora answer to "Are there any software tools or APIs that will allow you to create a 3D talking head that will generate correct lip movements when given a text stream?"

> http://stallios.com/

Since that Quora post I've come across a new player in the space, Stallios Mayca, apparently targeting the high-end => https://www.youtube.com/user/MaycaVideos/

[Art]

Nice but she can't hold a candle to Guile3D's Denise as far as a beautiful, lifelike character. But then again, beauty is rather subjective.

The lip-sync issue has plagued many would be developers in this digital avatar industry but I feel that soon, one day, they'll get it right and it will will work nicely on all platforms. There...that should help brighten everyone's day! Or something like that....

[Freddy]

Thanks Marcus, but I wasn't really looking for an end product - this is something I am doing for fun and to learn things. Which means I assemble or code it all by myself.

I agree with Art, that Stallios avatar is nice, but the lip-sync is a bit iffy in my humble opinion.

[8pla.net]

I know the reason for publishing a trial run is to get some useful feedback.  It seems as if the audio track may have been rendered slightly behind the video track by a few tenths of a second.  By looking very closely, the lips may continue to move a fraction of a second after the speech stops.

[Art]

... By looking very closely, the lips may continue to move a fraction of a second after the speech stops.

Yeah...I know of some politicians who exhibit similar behavior! 

[Freddy]

Thanks for the observation 8pla. I just tried talking and my lips might still be moving for that fraction of time 

But I know what you mean, I am at the mercy of the tools I play with.

[8pla.net]

Goes without saying, your trail-run render looks fantastic!  Good point!  Perhaps fine tuning audio skew in a 3D avatar may be more an art than a science.  Please share with me your reasoning.  In other words, and in your opinion, why do the phoneme sounds (the voice) reach the listener before the  physical lip postures are complete (mouth finishes moving) ?   Maybe the speed of sound is a factor, or something?

[Freddy]

I don't think it's the speed of sound really over such a short distance. It's probably because my PC can only process the graphical changes at that speed.

I use a decay setting, or in other words I cause it to slowly taper off a morph (lip shape). Perhaps I need to shorten that decay so that the mouth shapes finish more quickly.

Like you observed; it's a trial run, I ain't even remotely finished yet...