MIDI / transcriptionGPU workloadscore-informed-amt · Source score-informed-amt · Weights Score-HPT · Paper

Piano velocity refinement · Score-HPT

Upload piano audio and a time-aligned piano MIDI to estimate note velocities. Pitches, timing, tracks, tempo maps and pedal events are preserved. Notes are not added or aligned automatically.

15 credits/useFile inputMax 100 MBEstimated 30-2400 seconds

Input

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

Supported formats

mp3wavflacoggm4aaacaifaiffmp4movmkv

Output

MIDI

Use case

Upload piano audio and a time-aligned piano MIDI to estimate note velocities. Pitches, timing, tracks, tempo maps and pedal events are preserved. Notes are not added or aligned automatically.

Models, parameters, and sources

Score-HPTScore-HPT · Table 1 · MAESTRO v3 testHPTScore-HPT · onsetScore-HPT · onset + frameScore-HPT · onset + frame_exScore-HPT · Table 2 · SMDDiffVelFiLM ConvFiLM U-NetScore-HPPNetScore-DynEst

Score-HPT

zhanh-he/score-informed-amt Score-HPT Score-HPT

Paper Tables 1–2: MAESTRO-only training; MAE and STD measure velocity on the 0–127 scale at known note onsets. Lower is better. This tool uses onset + frame conditioning and samples at onsets by default. Paper scores are not site measurements.

Test setting

Paper Tables 1–2: MAESTRO-only training; MAE and STD measure velocity on the 0–127 scale at known note onsets. Lower is better. This tool uses onset + frame conditioning and samples at onsets by default. Paper scores are not site measurements.

Score-HPT

Official / repositories

zhanh-he/score-informed-amt Score-HPT

Papers / technical notes

Score-HPT

Parameter guide

Paper Tables 1–2: MAESTRO-only training; MAE and STD measure velocity on the 0–127 scale at known note onsets. Lower is better. This tool uses onset + frame conditioning and samples at onsets by default. Paper scores are not site measurements.

Workflow

1Upload a piano recording and time-aligned piano MIDI, then choose a velocity estimation method.
2Download the MIDI with updated velocities; notes and their timing remain unchanged.

Models in this field

2 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

The selected model has no verified matching public result and has not been ranked.

Score-HPT · Table 2 · SMD

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 2 / source rows 11

Not ranked: 9

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

DiffVel · Score-Inf. Methods · FiLM Conv · Score-Inf. Methods · FiLM U-Net · Score-Inf. Methods · Score-HPPNet · Proposed · Score-DynEst · Proposed · HPT · AMT Systems · HPPNet (no CQT) · AMT Systems · DynEst (no BSSL) · AMT Systems · hFT-Transformer · AMT Systems

Rank by metric

Score-HPT · Table 2 · SMD
Metric rank	Model	Test setting	SMD dataset
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
1	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	Proposed	8.0	6.9	5.0 M
2	Transkun v2 Public weights · Inference code benchmarks/Transkun/transkun/pretrained/2.0.pt Inference codeWeight license not declared	AMT Systems	13.3	11.9	13.6 M

Original records (2)

Score-HPT · Table 2 · SMD ↗Checked: 2026-10-25

All evaluation sources (2)

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

Score-HPT · Table 1 · MAESTRO v3 test ↗
Deployable entries 1 / source rows 4
Not ranked: 3
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
HPT · audio · Score-HPT · + onset · audio · Score-HPT · + onset · + frame ex · audio
Score-HPT · Table 2 · SMD ↗
Deployable entries 2 / source rows 11
Not ranked: 9
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
DiffVel · Score-Inf. Methods · FiLM Conv · Score-Inf. Methods · FiLM U-Net · Score-Inf. Methods · Score-HPPNet · Proposed · Score-DynEst · Proposed · HPT · AMT Systems · HPPNet (no CQT) · AMT Systems · DynEst (no BSSL) · AMT Systems · hFT-Transformer · AMT Systems

All original results and sources

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

Score-HPT · Table 1 · MAESTRO v3 test

Deployable entries 1 / source rows 4

Not ranked: 3

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

HPT · audio · Score-HPT · + onset · audio · Score-HPT · + onset · + frame ex · audio

Rank by metric

Score-HPT · Table 1 · MAESTRO v3 test
Metric rank	Model	Test setting	MAESTRO test set
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
-	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	+ onset · + frame · audio	4.0	4.7	5.0 M

Original records (1)

Score-HPT · Table 1 · MAESTRO v3 test ↗Checked: 2026-10-25

Score-HPT · Table 2 · SMD

Deployable entries 2 / source rows 11

Not ranked: 9

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

Rank by metric

Score-HPT · Table 2 · SMD
Metric rank	Model	Test setting	SMD dataset
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
1	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	Proposed	8.0	6.9	5.0 M
2	Transkun v2 Public weights · Inference code benchmarks/Transkun/transkun/pretrained/2.0.pt Inference codeWeight license not declared	AMT Systems	13.3	11.9	13.6 M

Original records (2)

Score-HPT · Table 2 · SMD ↗Checked: 2026-10-25

MIDI / transcriptionGPU workloadscore-informed-amt · Source score-informed-amt · Weights Score-HPT · Paper

Piano velocity refinement · Score-HPT

Upload piano audio and a time-aligned piano MIDI to estimate note velocities. Pitches, timing, tracks, tempo maps and pedal events are preserved. Notes are not added or aligned automatically.

15 credits/useFile inputMax 100 MBEstimated 30-2400 seconds

Input

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

Supported formats

mp3wavflacoggm4aaacaifaiffmp4movmkv

Output

MIDI

Use case

Upload piano audio and a time-aligned piano MIDI to estimate note velocities. Pitches, timing, tracks, tempo maps and pedal events are preserved. Notes are not added or aligned automatically.

Models, parameters, and sources

Score-HPT

zhanh-he/score-informed-amt Score-HPT Score-HPT

Paper Tables 1–2: MAESTRO-only training; MAE and STD measure velocity on the 0–127 scale at known note onsets. Lower is better. This tool uses onset + frame conditioning and samples at onsets by default. Paper scores are not site measurements.

Parameter guide

Paper Tables 1–2: MAESTRO-only training; MAE and STD measure velocity on the 0–127 scale at known note onsets. Lower is better. This tool uses onset + frame conditioning and samples at onsets by default. Paper scores are not site measurements.

Workflow

1Upload a piano recording and time-aligned piano MIDI, then choose a velocity estimation method.
2Download the MIDI with updated velocities; notes and their timing remain unchanged.

Models in this field

2 entries

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

The selected model has no verified matching public result and has not been ranked.

Score-HPT · Table 2 · SMD

Deployable entries 2 / source rows 11

Not ranked: 9

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

Rank by metric

Score-HPT · Table 2 · SMD
Metric rank	Model	Test setting	SMD dataset
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
1	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	Proposed	8.0	6.9	5.0 M
2	Transkun v2 Public weights · Inference code benchmarks/Transkun/transkun/pretrained/2.0.pt Inference codeWeight license not declared	AMT Systems	13.3	11.9	13.6 M

Original records (2)

Score-HPT · Table 2 · SMD ↗Checked: 2026-10-25

All evaluation sources (2)

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

Score-HPT · Table 1 · MAESTRO v3 test ↗
Deployable entries 1 / source rows 4
Not ranked: 3
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
HPT · audio · Score-HPT · + onset · audio · Score-HPT · + onset · + frame ex · audio
Score-HPT · Table 2 · SMD ↗
Deployable entries 2 / source rows 11
Not ranked: 9
A complete public inference package for these experiments has not been verified. Other weights with the same model name do not qualify.
DiffVel · Score-Inf. Methods · FiLM Conv · Score-Inf. Methods · FiLM U-Net · Score-Inf. Methods · Score-HPPNet · Proposed · Score-DynEst · Proposed · HPT · AMT Systems · HPPNet (no CQT) · AMT Systems · DynEst (no BSSL) · AMT Systems · hFT-Transformer · AMT Systems

All original results and sources

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

Score-HPT · Table 1 · MAESTRO v3 test

Deployable entries 1 / source rows 4

Not ranked: 3

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

HPT · audio · Score-HPT · + onset · audio · Score-HPT · + onset · + frame ex · audio

Rank by metric

Score-HPT · Table 1 · MAESTRO v3 test
Metric rank	Model	Test setting	MAESTRO test set
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
-	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	+ onset · + frame · audio	4.0	4.7	5.0 M

Original records (1)

Score-HPT · Table 1 · MAESTRO v3 test ↗Checked: 2026-10-25

Score-HPT · Table 2 · SMD

Deployable entries 2 / source rows 11

Not ranked: 9

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

Rank by metric

Score-HPT · Table 2 · SMD
Metric rank	Model	Test setting	SMD dataset
Metric rank	Model	Test setting	MAE ↓	STD ↓	# Params
1	Score-HPT Public weights · Inference code 100000_iterations.pth Inference codeWeight license not declared	Proposed	8.0	6.9	5.0 M
2	Transkun v2 Public weights · Inference code benchmarks/Transkun/transkun/pretrained/2.0.pt Inference codeWeight license not declared	AMT Systems	13.3	11.9	13.6 M

Original records (2)

Score-HPT · Table 2 · SMD ↗Checked: 2026-10-25

Piano velocity refinement · Score-HPT

Models, parameters, and sources

Score-HPT

Test setting

Official / repositories

Papers / technical notes

Parameter guide

Workflow

Models in this field

Score-HPT · Table 2 · SMD

Score-HPT · Table 1 · MAESTRO v3 test

Score-HPT · Table 2 · SMD

Piano velocity refinement · Score-HPT

Models, parameters, and sources

Score-HPT

Test setting

Official / repositories

Papers / technical notes

Parameter guide

Workflow

Models in this field

Score-HPT · Table 2 · SMD

Score-HPT · Table 1 · MAESTRO v3 test

Score-HPT · Table 2 · SMD

Quick feedback