In August last year, we realized that there was no way to convert Sindhi audio into text. We tried hard to find a model that could do this, but there wasn't one. So, we embarked on a journey to collect audio-transcript pairs in any way possible and trained our own speech-to-text model for Sindhi.
You can try our model here: https://huggingface.co/spaces/fahadqazi/Sindhi-Speech-to-Text