I didn’t fully understand HDVM yet.

Is it on by default if not where is the setting? Do all AI voicebanks have them? Is it good to have or bad?

“HDVM” is just a term for improvements made to the engine to produce more realistic and dynamic vocals.

It’s not something that can be enabled or disabled completely, it’s just a function the engine uses when rendering. In some ways you can reduce or increase the effects of it by changing the “expressiveness” slider in the Timbre AI Retakes panel (Pro edition only).

When it was introduced in 1.7.0 it was unstable and caused odd results, but as of 1.7.1 those issues have been fixed.

1 Like

HDVM is an under-the-hood update that represents two things: an underlying update to the editor software itself, and an iteration of the song database version. As such, it is not a feature that needs to be enabled; when you download version 1.7.0 or higher of the editor and the corresponding AI voice database, it already includes HDVM’s work product.

HDVM was originally designed to provide more variation and possibilities for AI vocals, which led to the introduction of AI Retakes. With AI Retakes, AI vocals can respond to the same verse and melody with different singing ways, and the creator can select the one that best fits the needs of the song and merge it into the song. This is a unique restoration of each real-life singer’s vocal, and a digital reproduction of the traditional music production process.