BiaB already offers access to Sinsy, and it doesn't seem to have caught the imagination of users. It's currently one of the most popular approaches for vocal synthesis.
Hidden Markov Model: This is what Sinsy uses.This requires quite a bit of work to create the database of samples needed. It's quite similar to how BiaB currently does RealTracks. Concatenation: This is the approach that Vocaloid and UTAU take.I think that licensing SoftVoice is probably a cost-effective approach, although there are some issues, such as running on a Mac. Synthesis: This is the old-school approach, and something that I've been working on, and continuing to work with.
There are basically three different approaches that can be taken with singing vocal synthesis: As someone who uses Vocaloids and is actively developing a singing vocal synthesis program, I've got my own guesses, but they're not inside information, so take this with a large grain of salt.
0 kommentar(er)
