Dunder_Intercalaris+ is nearing 150k steps, which means it is time to post some demos!
He has no low range and that will need to be fixed before he can be released. The UTAU engine noise is intentional, I feel like it adds to the voicebank, and this definitely isn’t because I don’t wanna clean all the input data to remove it. I also don’t think I can fix the thing with breaths because it’s not real singing data.
I’m running out of colab credits so for now I’m just gonna train it to 150k steps and then get more data for the voicebank based on my actual singing voice.