Does anybody else think the post-1.3.0 AI version of Saki AI (and possibly other AI voices) sounds muddy?

I mentioned recent in a comment that I have two versions of Synthesizer V installed, but haven’t talked about why.

I knew 1.3.0 would bring about big AI voice changes so I deliberately kept a copy of the old version along with its voices for comparison purposes. But no matter what I do, I keep thinking that the 1.3.0 voice update was terrible for Saki AI at the very least.

Here’s an example of what I mean. Here’s my attempt at doing the first half of the chorus of “Promise” by Kohmi Hirose (aka. the Geddan meme) with Saki AI ver. 104 in SynthV Studio Pro 1.2.4. I’ve uploaded it in WAV format to be sure that no compression artifacts exist, so it’s 1.4MB:

And here’s the exact same thing with Saki AI ver. 110 in SynthV Studio Pro 1.3.0:

To my ears, wearing headphones, the quality of the 1.3 voice sounds muddy (as if there are MP3 artifacts), and the breaths are noticeably worse, while the 1.2 voice sounds clean.

Has anyone else noticed this?


I also still use Saki AI 104. While I do think the gen3 update brought improvements in some areas, it also brought with it reduced flexibility and less user control. With newer versions of Saki it often feels like I have to fight the AI to achieve the desired sound.

It does sound muddy and weird. You’re absolutely right.