Guest
I sound kinda bad in this, cause it was a one take poorly cut down guide; Will be remaking with a good voiceover where I dont stutter over half the words sometime soon. But it works for now if you want to learn. I am still working on the full guide. It will go over vocal isolation, reverb removal (Reverb HQ / De-Echo models in UVR), etc. Linked below you will find two Google Docs I wrote, one for general use, and one for training your own voices. EDIT 6/11/2023: mangio-crepe is the better option over the official RVC crepe. I would recommend you use that, but you can try the official crepe if you want. mangio-crepe is the same thing as the option I cover in this guide. COLAB LINK: UVR DOWNLOAD (Isolate vocals with this): RVC v2 GENERAL SONG COVER GUIDE: RVC v2 VOICE MODEL TRAINING GUIDE: LOCAL INSTALL GUIDE INSTEAD OF COLAB (if you have a decent GPU): AI HUB DISCORD LINK: PUBLIC ARCHIVE OF VOICES INCASE YOU CANT JOIN THE AI SERVER: For most people, with UltimateVocalRemover, Kim vocal 1 or Inst HQ 1/2 will be solid choices for ripping the song's vocals. Then Reverb HQ followed by De-Echo Normal, or De-Echo-DeReverb (when reverb and/or echoey vocals are involved). If Reverb HQ seemed to do nothing, just skip it and do De-Echo instead as a whole. (Also heard that applying Reverb HQ on the OG song before isolating the vocal track is cleaner sometimes but not always???) Plus, if you need to do harmonies, 5HP karaoke is usually best for isolating backing vocals; this is ideal so you can do them separately, otherwise the AI will glitch trying to capture two vocal tracks in one vocal layer. I also said 'the loud sections will be quiet' by accident at like 3:15, I was basically trying to explain the volume envelope setting controlled how much the volume matched certain parts of your song. So loud parts will still be loud, its just the quiet parts will actually be quieter, it allows for range. (Special thanks to the KFAD siivagunner artists for the Mr Krabs render used in the thumbnail)
