My current voice workflow for AI shorts, and how do you order lip sync vs dialogue?
This is what I do now, and I'm still adjusting it.
- One voice card per character before I write a line: timbre, pace, volume, verbal habits. Once it's written I don't touch it, because a mid-scene voice change is the fastest way to lose people.
- Lines get broken short, with pauses and stress marked on the page.
- Emotion goes in as what it sounds like. "Angry" becomes lower, slower, no lift at the end of the line.
- For key lines I generate several takes and pick the one that sounds most like that character. Voice settings stay fixed, only the pauses and emotion notes change. Then I match loudness across the scene.
- Dialogue, ambience and music sit on separate tracks. Whoever is speaking is on top. Ambience keeps running underneath at a slightly lower level, and music gets out of the way before the line starts. A voice that sounds dry is usually missing the room under it.
For lip sync, I currently render the picture with the dialogue first, fix the mouths, then swap the finished audio in with pace and pause positions unchanged. I don't know if that's the best order. Doing audio first seems to mean redoing the shot whenever a line changes.
How do you order lip sync and dialogue? And which slip is hardest to rescue once it's in: a voice that drifts between shots, or mouths that don't match?