MPECHO: A MELODY AND PHONEME-AWARE GENERATIVE FRAMEWORK FOR CONTROLLABLE COVER SONG GENERATION

Wei-Jaw Lee1, 2, 3, Hsuan-Yu Yeh1, 3, Ting-Yi Hu1, 2, Chih-Pin Tan1, 2, Fang-Duo Tsai1, 3, Yi-Hsuan Yang1, 2, 3
1GICE, National Taiwan University; 2AI CoRE, National Taiwan University; 3Taiwan AI Labs
ISMIR 2026

Abstract

Cover song generation (CSG) should preserve the melodic and linguistic content of a reference song while recreating the remaining musical components. The state-of-the-art model SongEcho utilizes F0 sequences and voiced/unvoiced (V/UV) tags for conditioning; however, implicit linguistic information from V/UV tags cannot guarantee lyric accuracy, leading to a high phoneme error rate (PER). Inspired by singing voice synthesis (SVS), we propose MPEcho, which integrates a phoneme encoder and a length regulator (LR) into the SongEcho framework. By providing explicit phoneme-level conditioning and precise temporal boundaries, MPEcho significantly reduces PER. To enable this, we developed Phonsa, a Whisper-based automatic transcription model that provides high precision phoneme-level annotations for singing voices, overcoming the scarcity of high-quality audio-phoneme pairs. Experimental results validate the effectiveness of Phonsa for alignment and MPEcho for end-to-end CSG.

Audio Samples Demo

This section shows a gt audio and the generated audio samples by the models reported in Table 4 in the paper. The model ID 1, 2, 3 are trained with M (melody) condition, P (phoneme) condition and M+P conditions respectively. The model ID 9 is the same model with ID 3 but use multi-condition guidance.

Text Prompt Lyrics GT ID: 1 ID: 2 ID: 3 ID: 9
slap bass, rhythmic, brass, soulful, upbeat, tight, funk 從今後就選擇沉默
選擇服從歲月如梭
bright, happy, pop, acoustic guitar, melodic, radio-friendly, clean vocals 若非愛的天才 還能不能盼未來
一點愛 就能讓地裂天開
能讓我 把傷感重新掩埋
等你吹散陰霾 活出我的神采
soul, emotional, keyboard, grooving, romantic, soft, r&b 她的表情會變的有點怪怪
可是我還是不明白
為什麼她不理不睬
難道要老土的說一聲
對面的女孩看過來
對面的女孩看過來
Won't you just give me a try
nylon guitar, festive, congas, rhythmic, passionate, dance, latin 我的手你的臉越分越開
記得溫度記得幸福人才能釋懷
如果快樂你忘了帶
想想我們相愛的平凡
那種快樂那種執著
我相信誰都不能取代
fuzzy guitar, moody, heavy, 90s style, grunge 我需要一個擁抱
如果你在身邊 就好
我需要一個訊號
是你想念有我的座標
把地圖縮得再小
也不能抵擋難相見的煎熬
我知道我們很好
不會被距離 打倒