search

JSALT2022CodeSwitchingASR / generating-code-switched-audio

11 stars 3 forks source link

does the generated audios belong to a specific speaker identity? #7

Open thivux opened 1 year ago

thivux commented 1 year ago

since I could not find the audio samples in this repo, I wonder if the generated audios belong to a specific speaker identity?

AmirHussein96 commented 8 months ago

In general, the answer is no. The audio generation is speaker independent, meaning each word is sampled from a bag containing all instances of the same word, which very likely come from different speakers. I will try to add some samples in the coming days.

thivux commented 7 months ago

@AmirHussein96 hi can you please upload the audio samples?