Hyungchan Yoon (윤형찬)

Question about auxiliary features (BAP)

1

Thank you for your great work! One question related to your another [paper](https://ieeexplore.ieee.org/document/10246854), have you ever tried replace BAP(3-dim) with MAP(21-dim) in SiFiGAN experiments? I wonder whether MAP outperforms BAP...

About MAS (monotonic alignment search)

8

I've also implemented an E2E system using a CFM prior (different flow matching architecture instead of 1D-Unet). Despite using the prior loss in Grad-TTS, alignment framework fails to converge. (Prior...