search

FredHutch / loqui-vc

Shiny app for Creating Automated Videos with Voice Cloning
0 stars 0 forks source link

Multiple voices to clone to make a blended "fake" voice #1

Open howardbaik opened 10 months ago

howardbaik commented 10 months ago

Multiple people could speak in the training audio that we supply and then see how well it does to blend them.

cc @carriewright11

howardbaik commented 10 months ago

Seems doable with https://tts.readthedocs.io/en/dev/models/xtts.html#multiple-references

howardbaik commented 9 months ago

Voice Fusion is available on Coqui Studio: https://coqui.ai/blog/tts/voice-fusion.

Demo: https://youtu.be/2bUBYXEtvvw?si=duxVtoQ23S-NuQy8

howardbaik commented 9 months ago

@carriewright11 I'm wondering what would be the use case for combining multiple voices? I can kinda see the benefit of combining voices of the same sex, but of opposite sexes? That sounds weird to me...

cansavvy commented 9 months ago

Sometimes voices that sound traditionally "feminine" or "masculine" can lead people to associate stereotypes and interact differently with the material. A blended voice may help avoid stereotypes.

Additionally many folks don't identify with female or male so having a voice that doesn't identify as either is representing those folks as well.

carriewright11 commented 9 months ago

it's also better for security reasons