Contact us!
If you encounter an error, please reach out to the Lipdub Team.
- Discord: Click on the <?> button in the lower right hand corner of the Lipdub app to join the discord community
- Email: Discord is preferred, but if not possible, reach out to the support team at support@lipdub.ai
Accessing Lipdub’s Discord server
Source audio is not required
When uploading the source videos in Step #1: "Upload Videos". The audio is NOT important. The audio only comes into play at Step #4: "Upload Audio & Generate" when you upload the dub tracks for each actor.
Therefore when uploading videos in Step #1: "Upload Media", the videos can be all of the above:
- Mute
- With the original language OR
- With the dub language
whichever is easiest for your process.
Upload training footage to training section
We recommend uploading footage that is exclusively for training, into the training section. This will keep your project organized!
Video footage that is uploaded to this “Additional Footage for Training” will not be Lip-synced.
Check out What is ‘Additional Footage for Training’ & How to capture it On-Set? to understand what we mean by Additional Footage for Training.
Expressivity Control
Currently there is no direct slider that increases or decreases expressivity in the result. However there is one thing you can do as a user to potentially boost the emotion in the result to get your desired emotion.
The AI tracks the face of the actor, and if the actor’s mouth is smiling as they are talking, the more and more footage that is provided when they are smiling and talking at the same time will increase the chances of a result that is also smiling while they talk.
DISCLAIMER: This is not a guaranteed to work for all cases & situations. This is one example of a workaround that may result in a more expressive output.
Serious emotion - more footage was uploaded of just this actor making a serious face while talking
Smiling emotion - more footage was uploaded of just this actor smiling face while talking
Check the ‘unassigned tracks’ on Label Actors step
Make sure you check the ‘unassigned track’ section on the label actors step!
There’s a chance that some face detections belonging to speaking actors have not be assigned correctly. If they are, they need to be re-assigned to the correct actor.
The following video shows a user checking the ‘unassigned tracks’ section and seeing face detections that belong to Adam. The tracks are then being correctly assigned to Adam