Phoneme Maps
Last updated
Last updated
Phoneme Maps are assets used by AutoSync to convert Phonemes between two formats. These can either be actual LipSync PhonemeSet assets, or an external phoneme set used by another library. For example, the Montreal Forced Aligner module's English language model outputs phonemes using the ARPABET phoneme set - in order to use this data, there is an included Phoneme Map to convert from this format into the Preston Blair set that is used in LipSync Pro by default. This Phoneme Map can be found in the AutoSync/Phoneme Maps folder.
Custom Phoneme Maps can be created either by duplicating and modifying an existing one, or from the Create Asset menu, under LipSync Pro/AutoSync/External Phoneme Map.
The top of the Inspector for a Phoneme Map allows you to set a Display Name, which is used for listing Phoneme Maps in some circumstances. Below that are Phoneme Set A and Phoneme Set B - these are the names of the Phoneme Sets this map converts between. If referring to a PhonemeSet asset, make sure this exactly matches the scripting name set there.
The rest of the Inspector is taken up with the Phoneme List. This is a simple list of all phonemes in Set A and their equivalent phoneme in Set B. Any phonemes not found in the list when mapping will be ignored. The same Phoneme Map can be used in both directions, but A -> B will be favoured when automatically matching using AutoSyncUtility.FindBestFitPhonemeMap. If the same label appears more than once when doing a lookup, the first instance will always be used. Because of this, maps will generally have a single intended direction and while multiple phonemes in either set can map to the same phoneme in the other, they will usually only have each label in Set A appear once.