3 ms·
What scenario would be considered "zero shot" voice cloning? At the very least, wouldn't you have to provide 1 sample? Which would make it "few shot" (if that
by cjonas 2y ago
What scenario would be considered "zero shot" voice cloning?
At the very least, wouldn't you have to provide 1 sample? Which would make it "few shot" (if that term really even makes sense in this context).
- IanCal 2y agoI think the key distinction is that there is no specific training data for that speaker. You can view the input as just the input voice to clone, not training examples. It would be more like training examples if you had to give it specific phrases.