Toward Part-Aware Choral Transcription with singing voice assignment

Hanyu Meng; Zhanhong He; Zixun Guo; Yaolong Ju

eess.ASOct 7, 2026

Toward Part-Aware Choral Transcription with singing voice assignment

Authors: Hanyu Meng, Zhanhong He, Zixun Guo, Yaolong Ju

Organizations: The University of New South Wales, Sydney, Australia · Great Bay University, Guangdong, China · The University of Western Australia, Perth, Australia · Center for Digital Music (C4DM), Queen Mary University of London, London, United Kingdom

Abstract

Single-instrument automatic music transcription (AMT) has advanced substantially, yet choral applications require soprano, alto, tenor, and bass (SATB) to be transcribed as separate parts. Recent note-level choral AMT instead produces a single merged note track, limiting rehearsal, education, and score reconstruction. To address this limitation, we introduce Part-aware Choral Transcription (PawCT), to our knowledge the first end-to-end neural framework that identifies active SATB parts from choral audio and transcribes each into a separate note-level track. PawCT combines part-specific onset, offset, and frame prediction with part-presence estimation, union-level supervision, and structured training targets using a range prior (RP) based on SATB pitch ranges and its ordered-continuity (OC) extension, which adds within-part melodic continuity and cross-part pitch ordering. On YouChorale, PawCT-RP-OC achieves a macro part-aware note F1 of 0.225 at a 50-ms onset tolerance, outperforming an adapted choral baseline (0.165) by 36.4% relative and a two-stage post-hoc assignment pipeline (0.175). Its part-agnostic variant, PagCT, achieves a 50-ms onset F1 of 0.382, compared with 0.237 for the previous state-of-the-art choral AMT model. Cross-dataset evaluations on CSD and Cantoria further assess performance under dataset shift. These results demonstrate the benefit of jointly modeling note transcription and vocal-part assignment. Code and demos are available at https://hanyu-meng.github.io/Paw_Choral_AMT_Demo/.

Figures & tables

Figure 1: Part-agnostic versus part-aware choral transcription. PagCT produces a single merged track, whereas PawCT jointly identifies active SATB parts (green check marks) and transcribes them into separate note-level MIDI tracks.

Figure 2: Two strategies for part-aware choral transcription considered in this paper. (a) The proposed PawCT jointly transcribes mixed choral audio into SATB-specific note tracks. (b)–(c) The two-stage alternative first uses PagCT to produce a merged note track and then applies Post-VA, a symbolic voice-assignment model, to assign each decoded note to a vocal part.

Model	#Params	Frame			Onset (50 ms)			Onset (100 ms)
Model	#Params	Precision	Recall	F1	Precision	Recall	F1	Precision	Recall	F1
Onsets and Frames [ 12 ]	6.79M	0.806	0.326	0.428	0.450	0.178	0.242	0.688	0.248	0.344
MT3 [ 11 ]	44.7M	0.590	0.243	0.344	0.117	0.148	0.127	0.200	0.255	0.217
YourMT3+ [ 2 ]	45.8M	0.529	0.423	0.457	0.145	0.119	0.127	0.252	0.207	0.221
MuScriptor [ 28 ]	307M	0.720	0.745	0.730	0.086	0.083	0.084	0.236	0.228	0.231
Yu et al. [ 31 ]	4.23M	0.678	0.619	0.647	0.251	0.232	0.237	0.366	0.341	0.347

Table 1: Part-agnostic transcription on the YouChorale test set. Following [ 31 ] , we report frame and note precision, recall, and F1, with 50- and 100-ms onset tolerances for note metrics. “w/o Aug.” denotes no pitch-transposition augmentation. Blue rows indicate proposed systems; bold and underlined values denote the best and second-best results, respectively.

Model	Soprano		Alto		Tenor		Bass		Average		VA Rate (%)
Model	Frame	Note	Frame	Note	Frame	Note	Frame	Note	Frame	Note	Frame	Note
YourMT3+ ⋆ [ 2 ]	0.371	0.083	0.297	0.068	0.294	0.073	0.371	0.093	0.333	0.079	72.87	62.20
Yu et al. ⋆ [ 31 ]	0.528	0.183	0.451	0.151	0.436	0.138	0.560	0.186	0.494	0.165	76.35	69.62
PagCT + Post-VA	0.501	0.180	0.439	0.145	0.447	0.156	0.568	0.219	0.489	0.175	66.26	45.81
PawCT w/o union-loss	0.417	0.192	0.338	0.169	0.343	0.178	0.419	0.222	0.379	0.190	51.36	49.74
PawCT	0.530	0.228	0.430	0.191	0.429	0.197	0.552	0.252	0.485	0.217	65.72	56.81

Table 2: Part-aware SATB transcription performance on the YouChorale test set. For each voice part: Soprano, Alto, Tenor, and Bass, we report frame F1 and note F1 with a 50-ms onset tolerance, together with their macro-average. VA Rate is defined as

\mathrm{F1}_{\mathrm{avg}}/\mathrm{F1}_{\mathrm{agn}}\times 100\%

. All systems use pitch-transposition augmented training data. ⋆ denotes baselines adapted for part-aware transcription.

Dataset	Model	Soprano	Alto	Tenor	Bass	Avg.
CSD	PagCT + Post-VA	0.073	0.122	0.131	0.143	0.118
	PawCT	0.140	0.175	0.183	0.145	0.161
	PawCT-RP-OC	0.145	0.223	0.199	0.130	0.174
Cantoría	PagCT + Post-VA	0.179	0.115	0.085	0.084	0.115
	PawCT	0.240	0.239	0.275	0.328	0.271
	PawCT-RP-OC	0.329	0.273	0.300	0.348	0.313

Table 3: Cross-dataset evaluation of part-aware note F1 with a 50-ms onset tolerance on CSD and Cantoría. Models are trained only on YouChorale. Bold denotes the best result.

Explore similar work

VocalParse: Towards Unified and Scalable Singing Voice Transcription with Large Audio Language Models
May 6, 2026Yukun Chen, Tianrui Wang, Zhaoxi Mu +2Singing Voice Conversion Large Audio Language Models
Xi’an Jiaotong University · Nanyang Technological University · Tianjin University +2
SheetSage2: Coherent Lead-Sheet Transcription with Synthetic Supervision
Oct 4, 2026Junyan Jiang, Ruibin Yuan, Jiahao Pan +4Multi-Instrument Music Transcription
New York University · MBZUAI · ACE Studio +1
Snapping Matters: Context-Aware Onset Refinement for Automatic Music Transcription
Jun 10, 2026Abhirup Saha, Hans-Ulrich Berendes, Meinard Müller +1Multi-Instrument Music Transcription Speech-To-Text Alignment
International Audio Laboratories Erlangen

May 6, 2026, cs.SDJ/K move · Enter open · S save

VocalParse: Towards Unified and Scalable Singing Voice Transcription with Large Audio Language Models

Yukun Chen, Tianrui Wang, Zhaoxi Mu, Xinyu Yang+1

Xi’an Jiaotong University · Nanyang Technological University · Tianjin University+2

Singing Voice Conversion Large Audio Language Models Human Annotations

Open paper AskPDF arXiv

High-quality singing annotations are fundamental to modern Singing Voice Synthesis (SVS) systems. However, obtaining these annotations at scale through manual labeling is unrealistic due to the substantial labor and musical expertise required, making automatic annotation highly necessary. Despite their utility, current automatic transcription systems face significant challenges: they often rely on complex multi-stage pipelines, struggle to recover text-note alignments, and exhibit poor generalization to out-of-distribution (OOD) singing data. To alleviate these issues, we present VocalParse, a unified singing voice transcription (SVT) model built upon a Large Audio Language Model (LALM). Specifically, our novel contribution is to introduce an interleaved prompting formulation that jointly models lyrics, melody, and word-note correspondence, yielding a generated sequence that directly maps to a structured musical score. Furthermore, we propose a Chain-of-Thought (CoT) style prompting strategy, which decodes lyrics first as a semantic scaffold, significantly mitigating the context disruption problem while preserving the structural benefits of interleaved generation. Experiments demonstrate that VocalParse achieves state-of-the-art SVT performance on multiple singing datasets. The source code and checkpoint are available at https://github.com/pymaster17/VocalParse.

Abstract

Explore similar work

VocalParse: Towards Unified and Scalable Singing Voice Transcription with Large Audio Language Models

SheetSage2: Coherent Lead-Sheet Transcription with Synthetic Supervision

Snapping Matters: Context-Aware Onset Refinement for Automatic Music Transcription