The instant voice cloning feature on the platform lets you clone voices from very short audio samples (the audio needs to be of a single speaker, be ≈ 1 minute long and contain no background noise or music), but no model training is involved, hence the limit on quality.
We are also working on the voice-to-voice/voice conversion feature that will allow you to convert any voice into any other voice, those will come to Voice Air in the new year!!