MeanVC2 Released for Real-Time Voice Conversion

·2026.08.21 11:03

Key point

ASLP-lab has released MeanVC2, a real-time voice conversion model supporting cross-gender and cross-language conversion.

Details

ASLP-lab has released a new streaming voice conversion model, MeanVC2.

This model supports cross-gender and cross-language voice conversion, and can achieve 3x realtime processing on CPU environments using audio.cpp.

The model is accessible via the model card published on Hugging Face. The noise observed in the demo video has been confirmed to result from the demo engineering process rather than a quality issue with the model itself.

huggingface open-source audio-ai voice-conversion real-time-inference

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.