2.0 Pre release
Pre-release
Pre-release
CNChTu
released this
31 Jan 05:23
·
142 commits
to v2.0_dev
since this release
Diffusion SVC v2.0 is coming soon.
This model is a combination of NaiveV2, NaiveV2Diff, and Vocoder.
NaiveV2 and NaiveV2Diff is a cascaded training LYNXNet front stage and LYNXNet diffusion model.
They are extremely small in size and highly efficient.
You can train such a model by using configs/config_naivev2diff_comb.yaml and combining them with a fine-tuning vocoder using combo.py.
fine-tuning vocoder :https://github.com/openvpi/SingingVocoders