Sonorus OmniVoice AudioVAE

This repository contains the VAE-only GGUF used by Sonorus and the voxcpm2-upscaler branch of omnivoice.cpp to convert native OmniVoice output to 48 kHz audio.

File

File	Size	SHA-256
`voxcpm2-audiovae-f16.gguf`	187,868,032 bytes	`a5fb091c0a95172bdee2ee7230335dac7d3dc318d77ca100f095d023cabd5d97`

The file was converted from the official OpenBMB/VoxCPM2 audiovae.pth checkpoint using tools/convert-voxcpm2-audiovae.py from the omnivoice.cpp fork. The original model and weights are Apache-2.0. Conversion provenance is also embedded in the GGUF metadata.

This is only the AudioVAE. The OmniVoice language model and tokenizer GGUFs remain hosted in Serveurperso/OmniVoice-GGUF.

Downloads last month: 20

GGUF

Model size

93.8M params

Architecture

voxcpm2-audiovae

Hardware compatibility

16-bit

Model tree for Jrjy3/sonorus-omnivoice

Base model

openbmb/VoxCPM2

Quantized

(10)

this model