Female speaker "agastya" generating male audio note.

#2
by Mayanksa - opened

First of all thank you for this great model!

I noticed that for a shorter sentence, the speaker named "agastya" produced a male-sounding voice. Here is an example of the audio output:

English example which produced male audio

text_english = "Produce intelligence to generate human voices"
audio = generate_speech(text_english, speaker="agastya")
sf.write("output_english_agastya.wav", audio, 24000)

Yes , there are some other issues with this model , it hallucinates while using some commas in the text , dont think this can be used in production

Maya Research org

Early checkpointed model. Punctuations, numbers, names are what we are focusing on next. Also agastya is a male. Check it out in MayaResearch.ai/playground

bharathkumarK changed discussion status to closed

Agastya is sometimes speaking in Male and sometimes in female voice . How is it possible technically , just curious to understand the problem

Sign up or log in to comment