What is MuScriptor Multi-Instrument Transcription?

Turn a mixed recording into multi-instrument MIDI with MuScriptor-large.; Supports inputs in formats such as mp3 / wav / flac

Who is MuScriptor Multi-Instrument Transcription for?

MIDI / transcription Turn a mixed recording into multi-instrument MIDI with MuScriptor-large.; Supports inputs in formats such as mp3 / wav / flac

What tasks can MuScriptor Multi-Instrument Transcription help with?

MIDI / transcription; Turn a mixed recording into multi-instrument MIDI with MuScriptor-large.; Supports inputs in formats such as mp3 / wav / flac; Check whether the supported input formats match your material; Confirm whether the output fits your current workflow; Weigh runtime and credits to decide whether batch processing makes sense

Which model and keyword entities are connected with MuScriptor Multi-Instrument Transcription?

MuScriptor-large, MuScriptor source (pinned), MuScriptor-large weights (pinned), MuScriptor paper, MuScriptor Table 1, D_Test / CFG=1, Paper Table 1, YourMT3+

MuScriptor Multi-Instrument Transcription

MuScriptor-large

This route exposes only the flagship MuScriptor-large checkpoint; there is no small, medium, or quantized fallback.

MuScriptor source (pinned)MuScriptor-large weights (pinned)MuScriptor paper

Exact identity: 1,366,399,488 F32 parameters; 5,465,642,136-byte safetensors; revision 8809fdfbed2affa7ade94a7059e746e3880720e7.
Decoder-only Transformer with 48 layers, width 1536, and 24 heads; 16 kHz mono, 5-second chunks, and 512-bin mel at 100 Hz.
Outputs Type-1 multi-track MIDI; 35 UI labels map to 36 learned groups. Velocity remains fixed at 100. Processing automatically detects and writes the source BPM. After completion, playback speed and BPM can be adjusted in the preview control, and downloads use the control's current setting. Source tempo changes are not reconstructed.

Benchmark chart: Author-reported MuScriptor-large comparison / MuScriptor vs YourMT3+ across eight public datasets / MuScriptor model-scale ablation

Open-source deployable scope: these tables include only rows with verified downloadable weights or configurations and a local deployment path. SOTA and Best are relative to the verified rows shown, not hosted/API-only or commercial systems and not a universal leaderboard.

Author-reported MuScriptor-large comparison

MuScriptor Table 1

On the same private D_Test, the paper compares the final CFG=1 model with YourMT3+. The results are not independently reproduced and cannot be compared directly with MIROS or other evaluation protocols.

Evaluation condition	Model	Multi-instrument F1 (%)	Detailed metrics
D_Test / CFG=1	MuScriptor-largeTelkNetPaper Table 1	48.20SOTA	Onset F1 (%)60.40BestFrame F1 (%)73.30BestOffset F1 (%)49.00BestDrums F1 (%)50.20Best
D_Test / YPTF.MoE+Multi (noPS)	YourMT3+Paper Table 1	21.90	Onset F1 (%)32.52Frame F1 (%)45.54Offset F1 (%)17.79Drums F1 (%)41.40

MuScriptor vs YourMT3+ across eight public datasets

MuScriptor Table 2

Paper Table 2 evaluates the two released, locally deployable models on eight datasets excluded from both training sets. SOTA and Best are calculated within each dataset; MIROS is not inserted because the paper reports no row under this protocol.

Evaluation condition	Model	Multi-instrument F1 (%)	Detailed metrics
Bach10	YourMT3+Paper Table 2	26.4	Onset F1 (%)59.8BestFrame F1 (%)66.0Offset F1 (%)48.0Best
Bach10	MuScriptor-largeTelkNetPaper Table 2	34.7SOTA	Onset F1 (%)43.1Frame F1 (%)85.0BestOffset F1 (%)36.0
Dagstuhl ChoirSet	YourMT3+Paper Table 2	2.6	Onset F1 (%)22.3BestFrame F1 (%)51.0Offset F1 (%)10.8
Dagstuhl ChoirSet	MuScriptor-largeTelkNetPaper Table 2	11.5SOTA	Onset F1 (%)14.4Frame F1 (%)80.7BestOffset F1 (%)11.5Best
PHENICX-Anechoic	YourMT3+Paper Table 2	12.2	Onset F1 (%)56.7BestFrame F1 (%)58.9Offset F1 (%)18.7
PHENICX-Anechoic	MuScriptor-largeTelkNetPaper Table 2	25.7SOTA	Onset F1 (%)56.1Frame F1 (%)74.6BestOffset F1 (%)32.6Best
RWC-P	YourMT3+Paper Table 2	19.1	Onset F1 (%)36.1Frame F1 (%)51.6Offset F1 (%)20.6Drums F1 (%)36.5
RWC-P	MuScriptor-largeTelkNetPaper Table 2	25.6SOTA	Onset F1 (%)46.1BestFrame F1 (%)61.2BestOffset F1 (%)25.6BestDrums F1 (%)42.1Best
RWC-C	YourMT3+Paper Table 2	40.5SOTA	Onset F1 (%)71.7BestFrame F1 (%)71.3BestOffset F1 (%)44.3BestDrums F1 (%)8.7
RWC-C	MuScriptor-largeTelkNetPaper Table 2	36.0	Onset F1 (%)67.7Frame F1 (%)70.5Offset F1 (%)36.9Drums F1 (%)23.7Best
RWC-G	YourMT3+Paper Table 2	17.2	Onset F1 (%)36.9Frame F1 (%)49.4Offset F1 (%)20.5Drums F1 (%)25.6
RWC-G	MuScriptor-largeTelkNetPaper Table 2	23.7SOTA	Onset F1 (%)44.7BestFrame F1 (%)58.7BestOffset F1 (%)24.4BestDrums F1 (%)29.2Best
RWC-J	YourMT3+Paper Table 2	26.4	Onset F1 (%)52.9Frame F1 (%)57.2Offset F1 (%)31.1Drums F1 (%)30.6
RWC-J	MuScriptor-largeTelkNetPaper Table 2	31.8SOTA	Onset F1 (%)59.4BestFrame F1 (%)62.7BestOffset F1 (%)33.9BestDrums F1 (%)31.3Best
RWC-R	YourMT3+Paper Table 2	23.1SOTA	Onset F1 (%)46.1Frame F1 (%)61.5Offset F1 (%)28.6BestDrums F1 (%)51.3Best
RWC-R	MuScriptor-largeTelkNetPaper Table 2	20.3	Onset F1 (%)47.1BestFrame F1 (%)68.6BestOffset F1 (%)24.8Drums F1 (%)36.5

MuScriptor model-scale ablation

MuScriptor Table 4

Paper Table 4 uses D_Real-only training and the same private D_Test. Row sizes are paper labels, not exact public-checkpoint parameter counts.

Evaluation context	Model	Multi-instrument F1 (%)	Detailed metrics
D_Real only / conditioning on	MuScriptor 60MPaper Table 4	35.2	Onset F1 (%)47.7Frame F1 (%)65.7Offset F1 (%)35.3Drums F1 (%)39.8
D_Real only / conditioning on	MuScriptor 100MPaper Table 4	38.2	Onset F1 (%)51.2Frame F1 (%)67.2Offset F1 (%)38.7Drums F1 (%)41.5
D_Real only / conditioning on	MuScriptor 300MPaper Table 4	39.7	Onset F1 (%)52.4Frame F1 (%)68.0Offset F1 (%)40.3Drums F1 (%)42.0
D_Real only / conditioning on	MuScriptor 1.3BPaper Table 4	40.5	Onset F1 (%)53.2Frame F1 (%)68.7Offset F1 (%)41.0Drums F1 (%)42.5

Official / repositories

MuScriptor source (pinned)MuScriptor-large weights (pinned)MuScriptor paper

Papers / technical notes

No public paper link is listed; use the official repo or runtime implementation as the source.

MuScriptor Multi-Instrument Transcription

Models, parameters, and sources

MuScriptor-large

Benchmark chart: Author-reported MuScriptor-large comparison / MuScriptor vs YourMT3+ across eight public datasets / MuScriptor model-scale ablation

Author-reported MuScriptor-large comparison

MuScriptor vs YourMT3+ across eight public datasets

MuScriptor model-scale ablation

Official / repositories

Papers / technical notes

Parameter guide

Use cases

Workflow

Pre-flight checklist

Quick feedback