Spyro The Dragon (The Legend of Spyro: The Eternal Night)

Very compressed for the voice.

It also doesn't sound much like the voice in the game if I'm remembering TLOS correctly.

What did you train it on?

EasyGUI.

With Harvest or Crepe?

Crepe

can you tell me how to improve on it

use Mango crepe or if your dataset isn't good, use Harvest.

also use RVC v2 48k.

Harvest is good for iffy datasets that are noisy or/and compressed.

It covers up the noise and gets the best voice.

Mango Crepe is best for noise free datasets.

You can also clean up the dataset by isolating vocals and removing any background noise.

That also helps as well.

thank you

But if i train my model on harvest and i make inference with mangio crepe, it will sound good?

yes.

You can inference with something like Mango crepe even after you train with harvest.

It mainly focuses on wether you want better quality or less noise.

Noisy datasets are better with harvest and the safest bet.

Noise free are worth trying with Mango crepe.

After training Mango Crepe is best for quality, while Harvest is best if Mango Crepe can't keep the pitch right.

But what if i have only a bit of barely audible noise on the dataset?

Use audacity to do a noise gate.

it will remove the noise.

if that doesn't remove it, try rx 9 by izotope.

I don't have RX9 and i don't know how to use it

Did you try audacity's noise gate?

or noise reduction?

If you do noise reduction, you need to get the profile first by highlighting the area with noise, getting the profile, selecting all, then doing the reduction.

Thanks

no problem