Ulli | 2 years ago | 10 comments | 11 likes | 861 views
michiel, vincent, Decentralized and 8 others like this!
Hello my dear Friends,
this is a test with BluffTitler V 16 and a 3D singing woman.
I know it's not perfect yet.
Interesting, maybe there are some tries with various filtered audio settings. Anyway good work Ulli.
vincent, 2 years ago
I think it looks very convincing!
...or does this mean I'm very bad at reading lips? 😁
Maybe blinking the eyes helps making her look more alive.
michiel, 2 years ago
Bonjour Ulli,
Voilà un rapide exemple sur la chanson que tu as utilisé qui n'est pas parfait non plus, mais qui est peut être une petite avancée
Désolé mais je n’avais que mon avatar sous la main.
J’en avais discuté avec Michiel et je crois que le seul moyen pour avoir quelque chose d’acceptable et d’utiliser les effects MX par keyframe pour chaque mouvement des lèvres.
En pièce jointe le media file, ou se trouve la méthode que j’ai élaborée pour être le plus rapide possible :
• Dans le conteneur se trouvent une copie d’écran vidéo de la ligne de lecture sur mon LNE de l’audio et une copie en transparence du texte. Il faut synchroniser les feux fichier ensemble (éventuellement diviser en plusieurs partie le texte) . Peut-être que le texte transcrit en écriture phonétique serait encore plus pratique, mais je n’ai pas eu le temps…)
• Ensuite déplacer l’index de lecture, et en fonction appliquer le ou les effets Mx viseme correspondant à la diction
• Pour animer la tête, appliquer l’effet VJ au FX head avec des valeurs assez faibles.
Voilà, et bon courage :
drive.google.com...
Hello Ully,
Here's a quick example on the song you used which is not perfect either, but which is perhaps a small step forward
Sorry but I only had my avatar on hand.
I had discussed it with Michiel and I believe that the only way to have something acceptable in close-up is to use the MX effects by keyframe for each movement of the lips.
Attached is the media file, where is the method I developed to be as fast as possible:
• Inside the container is a video screenshot of the playback line on my LNE of the audio and a transparent copy of the text. It is necessary to synchronize the lights file together (eventually divide the text into several parts). Maybe the text transcribed in phonetic writing would be even more practical, but I did not have time…)
• Then move the reading index, and depending on it apply the Mx viseme effect(s) corresponding to the diction
• To animate the head, apply the VJ effect to the FX head with fairly low values.
Here it is, and good luck:
drive.google.com...
Alex-Raymond T., 2 years ago
Great job.
I have idea for when the vocals are not separated from the music. Record yourself singing the song with headphones on. Make the video with your recorded voice matched to the model. Render that, and just simply switch the audio in the video to the actual song.
My plan was to have Jesus singing the AC/DC song Highway to Hell. But I'm having trouble finding Jesus. Wait I mean...
Decentralized, 2 years ago
Me gusta esta muy bueno
gato@mo, 2 years ago
Thank you very much for your kindly comments, my dear Friends!
@ Alex
Thank you for your example. That was a lot of work for you. I will try it.
@ Michiel
Yes, you're right about eye movements!
@ Decentralized
Thanks for the tip!
Ulli, 2 years ago
