Lyrics-guided vocal transcription · STARS · TelkNet

MIDI / transcriptionGPU workloadSTARS · Source STARS · Weights STARS · Weights STARS · Chinese · Paper

Lyrics-guided vocal transcription

Requires the vocal audio plus lyrics, phonemes and a phoneme-to-word map. Exports note MIDI together with phoneme, word and vocal-technique timing; it is not an audio-only route.

25 credits/useFile inputMax 100 MBEstimated 60-2400 seconds

Input

Supports inputs in formats such as mp3 / wav / flac / ogg / m4a / aac / aif / aiff / mp4 / mov / mkv

Supported formats

mp3wavflacoggm4aaacaifaiffmp4movmkv

Output

MIDI / JSON

Use case

Edit notes, instrumentation and performance details.

Choose a model family

Lyrics-guided vocal transcription · STARSAvailableTranscribe notes and lyric timing from vocals, lyrics and phoneme information; choose by lyric language.

Choose a version or processing mode

Lyrics-guided vocal transcription · STARS · ChineseChinese lyric-guided singing transcription with note extraction and lyric timing.RecommendedAvailableLyrics-guided vocal transcription · STARS · BilingualChinese, English and mixed-lyric singing transcription with notes and lyric timing.Available

Choose a model for your material and purpose, and review the examples before submitting.

Cost, time, formats and parameters depend on the selected model.

Models, parameters, and sources

STARS · Chinese: Edit notes, instrumentation and performance details.

STARS · ChineseSTARS paper, Table 2STARS ChineseROSVOTVOCANOSTARS paper, Table 4STARS BilingualSTARS: A Unified Framework for Singing Transcription, Alignment, and Refined Style AnnotationSTARS · Bilingual

STARS · Chinese

STARS source (pinned)STARS weights (pinned)STARS paper

Enter a JSON array or whitespace-separated words. Contiguous Chinese text is split into characters.
Enter a JSON array or a space/comma-separated phoneme sequence using the selected STARS phone set.
Enter one zero-based word index per phoneme. The index count must equal the phoneme count.

STARS · Bilingual

STARS · Bilingual: Edit notes, instrumentation and performance details.

STARS source (pinned)STARS weights (pinned)STARS paper

Enter a JSON array or whitespace-separated words. Contiguous Chinese text is split into characters.
Enter a JSON array or a space/comma-separated phoneme sequence using the selected STARS phone set.
Enter one zero-based word index per phoneme. The index count must equal the phoneme count.

Paper and benchmark notes

STARS: A Unified Framework for Singing Transcription, Alignment, and Refined Style Annotation

ACL Findings - 2025

Paper Project demo

Official / repositories

STARS source (pinned)STARS weights (pinned)

Papers / technical notes

STARS paper

Parameter guide

Enter a JSON array or whitespace-separated words. Contiguous Chinese text is split into characters.
Enter a JSON array or a space/comma-separated phoneme sequence using the selected STARS phone set.
Enter one zero-based word index per phoneme. The index count must equal the phoneme count.

Workflow

1Upload vocals and supply lyrics, phonemes and phoneme-to-word indices.
2Generate MIDI and alignment annotations with STARS, then review note and lyric timing.

What to check

Enter a JSON array or whitespace-separated words. Contiguous Chinese text is split into characters.
Enter a JSON array or a space/comma-separated phoneme sequence using the selected STARS phone set.
Enter one zero-based word index per phoneme. The index count must equal the phoneme count.

Models in this field

3 entries

Includes public test results for models used here and elsewhere. The complete source table is shown by default; conditions and metrics can also be selected for comparison. Scores from different test conditions are not directly comparable.

Only records with verified public inference implementations and matching weights are shown; algorithms requiring no weights are also eligible. Closed services, unreleased weights and unmatched experiment versions are excluded. This is not an all-model leaderboard.

Public weights do not establish a fully open-source license. Expand the data for declared weight licenses and access conditions. Undeclared licenses are marked separately.

BenchmarkRanking category

STARS paper, Table 2

Choose a dataset to view all metrics for each model. Rows follow the official rank or selected metric. Bold marks each column’s best comparable value, including ties. Different evaluation protocols are ranked separately; missing scores are shown as -.

Deployable entries 3 / source rows 3

Values come from the linked public source. Row labels identify the test set or checkpoint; unreported scores are shown as -.

Rank by metric

STARS paper, Table 2
Metric rank	Model	COnPOff(F) ↑	RPA ↑
1	STARS (ours)Used on this site Public weights · Inference code stars_chinese/model_ckpt_steps_200000.ckpt Inference codeWeight license not declared	71.0	86.7
2	ROSVOT Public weights · Inference code checkpoints.zip · rosvot + rwbd + rmvpe Inference codeWeight license not declared	70.2	83.8
3	VOCANO Public weights · Inference code VOCANO segmentation Patch-CNN pitch Inference code`model.pt`Weight license not declared	50.2	76.6

Original records (3)

Markers identify models or components used here. Scores and settings come from the cited tests, not a new evaluation of the complete site pipeline.

STARS paper, Table 2 ↗Checked: 2026-10-25

All evaluation sources (2)

Datasets, versions and conditions remain separate, including sources with no eligible ranking rows.

STARS paper, Table 2 ↗
Deployable entries 3 / source rows 3
STARS paper, Table 4 ↗
Deployable entries 2 / source rows 6
Not ranked: 4
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
w/o CTC · w/o VQ · w/o MOE · Conv

All original results and sources

Scores apply to the listed dataset, metric and model version; results from different tests are not directly comparable.

STARS paper, Table 2

Deployable entries 3 / source rows 3

Values come from the linked public source. Row labels identify the test set or checkpoint; unreported scores are shown as -.

Rank by metric

STARS paper, Table 2
Metric rank	Model	COnPOff(F) ↑	RPA ↑
1	STARS (ours) Public weights · Inference code stars_chinese/model_ckpt_steps_200000.ckpt Inference codeWeight license not declared	71.0	86.7
2	ROSVOT Public weights · Inference code checkpoints.zip · rosvot + rwbd + rmvpe Inference codeWeight license not declared	70.2	83.8
3	VOCANO Public weights · Inference code VOCANO segmentation Patch-CNN pitch Inference code`model.pt`Weight license not declared	50.2	76.6

Original records (3)

STARS paper, Table 2 ↗Checked: 2026-10-25

STARS paper, Table 4

Deployable entries 2 / source rows 6

Not ranked: 4

A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.

w/o CTC · w/o VQ · w/o MOE · Conv

Values come from the linked public source. Row labels identify the test set or checkpoint; unreported scores are shown as -.

Rank by metric

STARS paper, Table 4
Metric rank	Model	BER ↓	IOU ↑	COnPOff(F) ↑	RPA ↑	T-F1 ↑	T-ACC ↑	S-ACC ↑
1	STARS Public weights · Inference code stars_chinese/model_ckpt_steps_200000.ckpt Inference codeWeight license not declared	18.6	80.9	71.0	86.7	79.9	91.5	68.0
2	Bilingual Public weights · Inference code stars_chinese_english_bilingual/model_ckpt_steps_300000.ckpt Inference codeWeight license not declared	19.4	75.6	68.1	86.2	76.8	91.4	70.4

Original records (2)

STARS paper, Table 4 ↗Checked: 2026-10-25

Lyrics-guided vocal transcription

Models, parameters, and sources

STARS · Chinese

STARS · Bilingual

Paper and benchmark notes

STARS: A Unified Framework for Singing Transcription, Alignment, and Refined Style Annotation

Official / repositories

Papers / technical notes

Parameter guide

Workflow

What to check

Models in this field

STARS paper, Table 2

STARS paper, Table 2

STARS paper, Table 4

Quick feedback