Dialogue, switch layers, Papagayo, et al

[toddwaddington]

Hello friends,

With all of the excitement around 13.5 (thank you, thank you, thank you! "Blessed Be"--oh, that's another thing), wondering about the future of lip syncing.

Best practices? Anyone still using Papagayo or do you find it's easier to use switch layers, smart bones? I think these are all personal preferences, but would love to know if there is new AI down the pike. For example, I do real estate photography and virtual tours. There is an amazing software called Vidnami. You write the copy and an AI of your choice voices it, populates with your photographs, and adds royalty free music.

With all of that technology available, I'm wondering how close we are to more accurate lip syncing tech. Just a geeky question. But would love to hear user experiences. Here is an example of a little commercial I made effortlessly. Yes, I could totally do it in my video editing software, but the idea of this *almost* undetectable voice filling auto filling is amazing. [youtube]https://youtu.be/Ccm5uE-VCbk[/youtube]

[Karl Toon]

13 did auto lip-sync without having to use Papagayo, but this has disappeared in 13.5 which is a shame as it was quite good and saved loads of time. Hopefully this feature will return at some point.

[Karl Toon]

<snip>

With all of that technology available, I'm wondering how close we are to more accurate lip syncing tech. Just a geeky question. But would love to hear user experiences. Here is an example of a little commercial I made effortlessly. Yes, I could totally do it in my video editing software, but the idea of this *almost* undetectable voice filling auto filling is amazing. [youtube]https://youtu.be/Ccm5uE-VCbk[/youtube]

Have just watched part of the commercial. Are you talking about lip-sync or text-to-speech?

[chucky]

I don't use Papayago, more effort than it's worth.

[cgrotke]

In 12.5 I liked assigning switch layers to an audio track, letting Moho guess, then doing some adjustments by hand to correct little things. Sometimes, for example, I'll want to hit a vowel harder than what Moho guessed. Or I'll make sure a mouth is closed completely when Moho thought it heard something... : )

This is good enough for me for most of my cartoon-y projects - if I were going for more realistic I might try something else.

What I like is that it is a very quick way to get "almost there" and the hand tweaking doesn't bother me.

(I've also used the speed control in the audio to speed up voices a bit sometimes. It's be nice to have finer controls over the tempo and pitch, but not bad as is.

[pixelblast]

Automated lip synching only used different sizes of open mouths and then "synched" that to the audio levels. That might be OK for Japanese Anime-Style animation, where they don't use different mouth shapes and the movement of the mouth hardly syncs to the audio, but if you want to proper lip **sync**, then you need proper mouth shapes for the different sounds.

Also when you enable interpolation for the mouth switch layer, it's best to do the actual switching manually, because the timing for this is different than using switch layers non-interpolated. Also in many situations you have to "skip" some mouth shapes, because the person talks to quickly, and then no automation will help you – only human judgement.

Doing it by hand in Moho is fairly quick, so it doesn't bother me.