A Different Definition of “Auto-Tune”

Here’s something that I thought was somewhat nifty: a system that generates tuning for Vocaloid based on parameters from a database of singing samples. Here’s some additional audio that contains a comparison between baseline untuned audio, manually tuned audio, and the audio with the generated parameters.

While the results section from the article reports that the version generated by the system tested the best in terms of expressiveness and naturalness, I still think that it soundsĀ  slightly off compared to the manually tuned version, but I do like it better than a lot of the results people often get with Vocalistener/Vocashifter/similar technologies.

Here’s the article, for those who are interested.