since the lumi model i made a year ago is starting to get stale, i decided to make a new one since she (allegedly) got a new recording setup.
trained on (almost) 16 minutes of audio data from her recent invaderversary stream
(mel-roformer fv7z + bs roformer resurrection (in some areas) + anvuew's room de-reverb + dialogue isolate (in some areas) + slight eq + de-click + de-plosive, resampled to 32k)
pitch extraction: rmvpe
steps: 18.9k
batch size: 7
pretrain: klm 4 hfg / 32k
act1: <#1249666913756381257>
huggingface: https://huggingface.co/sxndypz/rvc-v2-models/resolve/main/lumi_act2.zip
Tags: No tags available
Download Link: https://huggingface.co/sxndypz/rvc-v2-models/resolve/main/lumi_act2.zip Hosted on Hugging Face
Selected-sample labels, grouped by what they describe.
Pick a pitch, compare the Quality Score, then Run. Hold Ctrl or Cmd to select multiple pitches.
Nearby voices from generated samples. Pick a pitch in the Sample Deck to refresh the matches.