libwaifu-seed-vc

Seed-VC v2 voice conversion, exported for libwaifu: somebody speaking and a recording of somebody else in; the first one's words in the second one's voice out, at 22.05 kHz.

seed_vc.yaml -- the manifest; seed_vc-0000{1,2}-of-00002.safetensors -- 553 M parameters, float32.
Contents: v2/cfm_small.pth and v2/ar_base.pth from Plachta/Seed-VC; the bsq2048_light and bsq32_light ASTRAL tokenizers from Plachta/ASTRAL-quantization; the first 18 layers of facebook/hubert-large-ll60k; funasr/campplus; nvidia/bigvgan_v2_22khz_80band_256x.
Built by tools/seed_vc_exporter.py; see docs/seed_vc.md.

cargo run --release --features gpl --example convert -- seed_vc.yaml source.wav voice.wav out.wav

Licence

Seed-VC and ASTRAL are GPL-3.0, code and weights, and so is this redistribution. The runtime that reads it is behind libwaifu's gpl feature. The other components keep their own licences (HuBERT, CAMPPlus, BigVGAN -- see their repositories).

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ling0322/libwaifu-seed-vc

Base model

Plachta/ASTRAL-quantization

Finetuned

(1)

this model