Added new model card/alphabet file/license for the Yoloxóchitl Mixtec language (xty).
Dataset was modified from https://www.openslr.org/89/ by removing chunks of records where sox would crash when trying to figure out the number of audio channels. The data was then process by commonvoice utils (https://github.com/ftyers/commonvoice-utils) to convert into the mono audio to train with Coqui.
Did you find a way to filter out those chunks, or was it a matter of letting sox fail and then manually remove them? Do you have a script for (most of) this process?
Overview
Added new model card/alphabet file/license for the Yoloxóchitl Mixtec language (xty).
Dataset was modified from https://www.openslr.org/89/ by removing chunks of records where sox would crash when trying to figure out the number of audio channels. The data was then process by commonvoice utils (https://github.com/ftyers/commonvoice-utils) to convert into the mono audio to train with Coqui.
Model adding separately: STT-SLR89-XTY-0.1