Indexes of speaker embeddings
Hi! I want to make use of the different speakers in the dataset, but am unsure which indexes to use.
In my understanding each utterance of a speaker is represented with one index (for example 7306 is one utterance of an American woman), so you can load a speaker embedding with its index
speaker_embeddings = embeddings_dataset[7306]["xvector"]
But how would I find the "best" / "most understandable" index for each speaker? Is it simply an effort of trying multiple or is there some other technique? I am most interested in the following speakers:
bdl (US male)
rms (US male)
clb (US female)
Thanks a lot for your help!
Did you ever get an answer to this?
Not yet! But I found an example online which used the following files (which I am now using as well):
/cmu_us_bdl_arctic-wav-arctic_a0009
/cmu_us_clb_arctic-wav-arctic_a0144
/cmu_us_ksp_arctic-wav-arctic_b0087
/cmu_us_rms_arctic-wav-arctic_b0353
/cmu_us_slt_arctic-wav-arctic_a0508
Awesome, thank you
VOICE_INDEX = {
"awb": 0000, # Scottish male
"bdl": 1139, # US male
"clb": 2278, # US female
"jmk": 3417, # Canadian male
"ksp": 4556, # Indian male
"rms": 5695, # US male
"slt": 6834, # US female
"smp": 7306, # US female (exemplo de voz do modelo)
}