Datasets:
Tasks:
Automatic Speech Recognition
Formats:
webdataset
Size:
10M - 100M
ArXiv:
Tags:
croissant
DOI:
License:
can i use this dataset to finetune tts model?
#5
by
adityaakmalazhari
- opened
so i have a plan to finetune some voice cloner model like coqui xtts, cosy voice, f5-tts, etc for indonesian lang. Is this dataset good/clear enough for it? i have checked the refined train dataset by random sampling but the quality is almost 50:50 with the raw data. Some audio in refined data have some missing word between transcribe and audio, especially in last word of audio. For example, in audio i can only hear the speaker says 4 words but in the transcribe sometime 5 words.
CC @Sedrick99