What is disappointing to me is that RL just announced that the CC 3.43 update features compatibility between CC and Omniverse Audio2Face, which supports multiple languages!

So why can't that be implemented for AccuLips?

I think we should keep hammering on this until we get more than some non-committal answer. It is as if RL wants to chase the core filmmakers away from their products...:unsure:

---

I'm a Director artist , really not a Python coder.... I want to make movie with fewer mouse click.

There is a Python script from Mike Kelley to take the viseme output from Papagayo and then use that in iClone. That could perhaps be the basis for what you want to do.

See: https://forum.reallusion.com/427265/Improved-Papagayo-Script.

---

Mysterious duplicate post...:unsure:

---

I guess it will be a bit more work if your goal is to get the results back into iClone, but it should be possible with what commands are available in iClone's Python API.
I'm currently just working with English dialogue, but probably will have to deal with translations later.

To generate phonemes I have just followed the instructions on their website: Examples — Montreal Forced Aligner 2.0.0a22 documentation (montreal-forced-aligner.readthedocs.io)
- I have my dialogue organized in folders for each character
- I have for example LINE1.wav, and a corresponding LINE1.lab (this is just a text file with the dialogue written out, it's the same as the .txt file you can make for iClone).
- I run the command to generate the phonemes as shown in the example (but pointed at the directory where all my files are instead of the example dataset).

Now in the output directory, you have all these .TextGrid files with the phonemes and at what time they happen.
You would have to parse this file to extract the phonemes, and use it to place the visemes on the timeline.
In the iClone install folder there is a file "iClone 7\Resource\ICTextToSpeech\Dictionary\en.PhonemeVisemeMapping" where you can read which phonemes get turned into which visemes, for example both the "F" and "V" phonemes get turned into the "F_V" viseme.

I found that iClone is using a free, open source software called Montreal Forced Aligner to generate the phonemes, which is the main data I wanted to generate.
So, I have instead built my pipeline around this same software and skipped iClone entirely. Maybe I will find a use for iClone in the future

I started this thread last month about the montreal forced aligner :
https://forum.reallusion.com/Topic486876.aspx

Could you share/show what pipeline you use ? I would like to use french acculips.

Thank you

I am exporting a custom file format from the blend weight, bone transform, and phoneme values I am able to access on the viseme component.
I have the whole automation process fully working using the older lipsync function that only uses the audio file for input.
So the problem is not with the export, only with making the input be used in the desired way (AccuLips as opposed to the older, previously existing automatic lipsync function).
The only thing I am missing is the AccuLips feature which produces noticeably better lipsync.

@josh: So I assume your desired output is a MotionPlus file for each line?

Currently, that would be the only way to record the facial animation, AFAIK. We used to have .iTalk files, but I think that's old hat.

That said, there may be other options that I'm not aware of...:unsure:

---

The lines are split up and I have .wav files and .txt files for each of them.
Using the user interface, even if I spent 1 minute on each line it would take 2 weeks to get through them.