NSynth is a dataset of one shot instrumental notes, containing 305,979 musical notes with unique pitch, timbre and envelope. The sounds were collected from 1006 instruments from commercial sample libraries and are annotated based on their source (acoustic, electronic or synthetic), instrument family and sonic qualities. The instrument families used in the annotation are bass, brass, flute, guitar, keyboard, mallet, organ, reed, string, synth lead and vocal. Four second monophonic 16kHz audio snippets were generated (notes) for the instruments. Source: Data Augmentation for Instrument Classification Robust to Audio Effects Image Source: https://magenta.tensorflow.org/nsynth
The MAESTRO dataset contains over 200 hours of paired audio and MIDI recordings from ten years of International Piano-e-Competition. The MIDI data includes key strike velocities and sustain/sostenuto/una corda pedal positions. Audio and MIDI files are aligned with ∼3 ms accuracy and sliced to individual musical pieces, which are annotated with composer, title, and year of performance. Uncompressed audio is of CD quality or higher (44.1–48 kHz 16-bit PCM stereo). Source: https://magenta.tensorflow.org/datasets/maestro Image Source: https://www.researchgate.net/figure/Results-generated-with-the-MAESTRO-dataset_fig3_333392458
MusicCaps is a dataset composed of 5.5k music-text pairs, with rich text descriptions provided by human experts. For each 10-second music clip, MusicCaps provides: 1) A free-text caption consisting of four sentences on average, describing the music and 2) A list of music aspects, describing genre, mood, tempo, singer voices, instrumentation, dissonances, rhythm, etc. Source:MusicLM: Generating Music From Text
POP909 is a dataset which contains multiple versions of the piano arrangements of 909 popular songs created by professional musicians. The main body of the dataset contains the vocal melody, the lead instrument melody, and the piano accompaniment for each song in MIDI format, which are aligned to the original audio files. Furthermore, annotations are provided of tempo, beat, key, and chords, where the tempo curves are hand-labelled and others are done by MIR algorithms. Source: https://arxiv.org/pdf/2008.07142.pdf Image Source: https://arxiv.org/pdf/2008.07142.pdf
The Lakh MIDI dataset is a collection of 176,581 unique MIDI files, 45,129 of which have been matched and aligned to entries in the Million Song Dataset. Its goal is to facilitate large-scale music information retrieval, both symbolic (using the MIDI files alone) and audio content-based (using information extracted from the MIDI files as annotations for the matched audio files). Around 10% of all MIDI files include timestamped lyrics events with lyrics are often transcribed at the word, syllable or character level. LMD-full denotes the whole dataset. LMD-matched is the subset of LMD-full that consists of MIDI files matched with the Million Song Dataset entries. LMD-aligned contains all the files of LMD-matched, aligned to preview MP3s from the Million Song Dataset. A lakh is a unit of measure used in the Indian number system which signifies 100,000. Source: https://colinraffel.com/projects/lmd/ Audio Source: https://colinraffel.com/projects/lmd/
The JSB chorales are a set of short, four-voice pieces of music well-noted for their stylistic homogeneity. The chorales were originally composed by Johann Sebastian Bach in the 18th century. He wrote them by first taking pre-existing melodies from contemporary Lutheran hymns and then harmonising them to create the parts for the remaining three voices. The version of the dataset used canonically in representation learning contexts consists of 382 such chorales, with a train/validation/test split of 229, 76 and 77 samples respectively.
EMOPIA (pronounced ‘yee-mò-pi-uh’) dataset is a shared multi-modal (audio and MIDI) database focusing on perceived emotion in pop piano music, to facilitate research on various tasks related to music emotion. The dataset contains 1,087 music clips from 387 songs and clip-level emotion labels annotated by four dedicated annotators.
The Groove MIDI Dataset (GMD) is composed of 13.6 hours of aligned MIDI and (synthesized) audio of human-performed, tempo-aligned expressive drumming. The dataset contains 1,150 MIDI files and over 22,000 measures of drumming.
VGMIDI is a dataset of piano arrangements of video game soundtracks. It contains 200 MIDI pieces labeled according to emotion and 3,850 unlabeled pieces. Each labeled piece was annotated by 30 human subjects according to the Circumplex (valence-arousal) model of emotion using a custom [web tool](https://github.com/lucasnfe/adl-music annotation).
ASAP is a dataset of 222 digital musical scores aligned with 1068 performances (more than 92 hours) of Western classical piano music. The scores are provided as paired MusicXML files and quantized MIDI files, and the performances as paired MIDI files and partially as audio recordings. Scores and performances are aligned with downbeat, beat, time signature, and key signature annotations. Source: ASAP: a dataset of aligned scores and performances for piano transcription
The Lakh Pianoroll Dataset (LPD) is a collection of 174,154 multitrack pianorolls derived from the Lakh MIDI Dataset (LMD). ## Getting the dataset We provide multiple subsets and versions of the dataset (see here). The dataset is available here. ## Using LPD The multitrack pianorolls in LPD are stored in a special format for efficient I/O and to save space. We recommend to load the data with Pypianoroll (The dataset is created using Pypianoroll v0.3.0.). See here to learn how the data is stored and how to load the data properly. ## License Lakh Pianoroll Dataset is a derivative of Lakh MIDI Dataset by Colin Raffel, used under CC BY 4.0. Lakh Pianoroll Dataset is licensed under CC BY 4.0 by Hao-Wen Dong and Wen-Yi Hsiao. Please cite the following papers if you use Lakh Pianoroll Dataset in a published work. - Hao-Wen Dong, Wen-Yi Hsiao, Li-Chia Yang, and Yi-Hsuan Yang, "MuseGAN: Multi-track Sequential Generative Adversarial Networks for Symbolic Music Generation and Accompaniment," in Proceedings of the 32nd AAAI Conference on Artificial Intelligence (AAAI), 2018. - Colin Raffel, "Learning-Based Methods for Comparing Sequences, with Applications to Audio-to-MIDI Alignment and Matching," PhD Thesis, 2016. ## Related projects - MuseGAN - LeadSheetGAN
ATEPP is a dataset of expressive piano performances by virtuoso pianists. The dataset contains 11677 performances (~1000 hours) by 49 pianists and covers 1580 movements by 25 composers. All of the MIDI files in the dataset come from the piano transcription of existing audio recordings of piano performances. Scores in MusicXML format are also available for around half of the tracks. The dataset is organized and aligned by compositions and movements for comparative studies. Source: ATEPP: A Dataset of Automatically Transcribed Expressive Piano Performances
The Nintendo Entertainment System Music Database (NES-MDB) is a dataset intended for building automatic music composition systems for the NES audio synthesizer. It consists of 5278 songs from the soundtracks of 397 NES games. The dataset represents 296 unique composers, and the songs contain more than two million notes combined. It has file format options for MIDI, score and NLM (NES Language Modeling). Source: https://github.com/chrisdonahue/nesmdb Image Source: https://github.com/chrisdonahue/nesmdb Audio Source: https://github.com/chrisdonahue/nesmdb
The MidiCaps dataset [1] is a large-scale dataset of 168,385 midi music files with descriptive text captions, and a set of extracted musical features. The captions have been produced through a captioning pipeline incorporating MIR feature extraction and LLM Claude 3 to caption the data from extracted features with an in-context learning task. The framework used to extract the captions is available open source on github. The original MIDI files originate from the Lakh MIDI Dataset [2,3] and are creative commons licenced.
The MusicBench dataset is a music audio-text pair dataset that was designed for text-to-music generation purpose and released along with Mustango text-to-music model. MusicBench is based on the MusicCaps dataset, which it expands from 5,521 samples to 52,768 training and 400 test samples! Dataset Details MusicBench expands MusicCaps by: Including music features of chords, beats, tempo, and key that are extracted from the audio. Describing these music features using text templates and thus enhancing the original text prompts. Expanding the number of audio samples by performing musically meaningful augmentations: semitone pitch shifts, tempo changes, and volume changes. Train set size = 52,768 samples Test set size = 400 This dataset also includes FMACaps, which was used as a second test set.
The ADL Piano MIDI is a dataset of 11,086 piano pieces from different genres. This dataset is based on the Lakh MIDI dataset, which is a collection on 45,129 unique MIDI files that have been matched to entries in the Million Song Dataset. Most pieces in the Lakh MIDI dataset have multiple instruments, so for each file the authors of ADL Piano MIDI dataset extracted only the tracks with instruments from the "Piano Family" (MIDI program numbers 1-8). This process generated a total of 9,021 unique piano MIDI files. Theses 9,021 files were then combined with other approximately 2,065 files scraped from publicly-available sources on the internet. All the files in the final collection were de-duped according to their MD5 checksum. Source: ADL Piano MIDI
The Song Describer Dataset (SDD) contains ~1.1k captions for 706 permissively licensed music recordings. It is designed for use in evaluation of models that address music-and-language (M&L) tasks such as music captioning, text-to-music generation and music-language retrieval.
The Bach Doodle Dataset is composed of 21.6 million harmonizations submitted from the Bach Doodle. The dataset contains both metadata about the composition (such as the country of origin and feedback), as well as a MIDI of the user-entered melody and a MIDI of the generated harmonization. The dataset contains about 6 years of user entered music. Source: https://magenta.tensorflow.org/datasets/bach-doodle Image Source: https://magenta.tensorflow.org/datasets/bach-doodle
ComMU has 11,144 MIDI samples that consist of short note sequences created by professional composers with their corresponding 12 metadata. This dataset is designed for a new task, combinatorial music generation which generate diverse and high-quality music only with metadata through auto-regressive language model. Source: ComMU: Dataset for Combinatorial Music Generation Image Source: https://arxiv.org/pdf/2211.09385v1.pdf
dMelodies is dataset of simple 2-bar melodies generated using 9 independent latent factors of variation where each data point represents a unique melody based on the following constraints: - Each melody will correspond to a unique scale (major, minor, blues, etc.). - Each melody plays the arpeggios using the standard I-IV-V-I cadence chord pattern. - Bar 1 plays the first 2 chords (6 notes), Bar 2 plays the second 2 chords (6 notes). - Each played note is an 8th note. Source: https://github.com/ashispati/dmelodies_dataset
The FakeMusicCaps dataset contains total of 27605 10 seconds music tracks corresponding to almost 77 hours, generated using 5 different Text-To-Music (TTM) models. It is designed to be used as a starting dataset for the training and/or evaluation of models for the detection and attribution of synthetic music generated via TTM models. If you use this dataset, please cite our paper: Comanducci, L., Bestagini, P., and Tubaro, S., “FakeMusicCaps: a Dataset for Detection and Attribution of Synthetic Music Generated via Text-to-Music Models”, Art. no. arXiv:2409.10684, 2024. doi:10.48550/arXiv.2409.10684.
A MIDI dataset of 500 4-part chorales generated by the KS_Chorus algorithm, annotated with results from hundreds of listening test participants, with 500 further unannotated chorales.
The AIME dataset contains 6,000 audio tracks generated by 12 music generation models in addition to 500 tracks from MTG-Jamendo. The prompts used to generate music are combinations of representative and diverse tags from the MTG-Jamendo dataset.
Dataset Summary The dataset used to train and evaluate TunesFormer is collected from two sources: The Session and ABCnotation.com. The Session is a community website focused on Irish traditional music, while ABCnotation.com is a website that provides a standard for folk and traditional music notation in the form of ASCII text files. The combined dataset consists of 285,449 ABC tunes, with 99% (282,595) of the tunes used as the training set and the remaining 1% (2854) used as the evaluation set. Control codes are symbols that are added to the ABC notation representation to indicate the desired musical form of the generated melodies. We add the following control codes to each ABC tune in the dataset through an automated process to indicate its musical form: - Number of Bars (NB): controls the number of bars in a section of the melody. For example, users could specify that they want a section to contain 8 bars, and TunesFormer would generate a section that fits within that structure. It counts on the bar symbol |. - Number of Sections (NS): controls the number of sections in the entire melody. This can be used to create a sense of structure and coherence within the melody, as different sections can be used to create musical themes or motifs. It counts on several symbols that are commonly used in ABC notation and can be used to represent section boundaries: [|,||,|],|:,::, and :|. - Edit Distance Similarity (EDS): controls the similarity level between the current section and a previous section in the melody. To ensure consistency and standardization among the ABC tunes in the dataset, we first converted them all into MusicXML format and then re-converted them back into ABC notation. In order to focus solely on the musical content, we removed any natural language elements (such as titles, composers, and lyrics) and unnecessary information (such as reference numbers and sources).
Guitar-TECHS is a comprehensive dataset featuring a variety of guitar techniques, musical excerpts, chords, and scales. These elements are performed by diverse musicians across various recording settings. Guitar-TECHS incorporates recordings from two stereo microphones: an egocentric microphone positioned on the performer’s head and an exocentric microphone placed in front of the performer. It also includes direct input recordings and microphoned amplifier outputs, offering a wide spectrum of audio inputs and recording qualities. All signals and MIDI labels are properly synchronized. Its multi-perspective and multi-modal content makes Guitar-TECHS a valuable resource for advancing data-driven guitar research, and to develop robust guitar listening algorithms.
Jazz pianists often uniquely interpret jazz standards. Passages from these interpretations can be viewed as sections of variation. We manually extracted such variations from solo jazz piano performances. The JAZZVAR dataset is a collection of 502 pairs of Variation and Original MIDI segments. Each Variation in the dataset is accompanied by a corresponding Original segment containing the melody and chords from the original jazz standard. Our approach differs from many existing jazz datasets in the music information retrieval (MIR) community, which often focus on improvisation sections within jazz performances. In this paper, we outline the curation process for obtaining and sorting the repertoire, the pipeline for creating the Original and Variation pairs, and our analysis of the dataset. We also introduce a new generative music task, Music Overpainting, and present a baseline Transformer model trained on the JAZZVAR dataset for this task. Other potential applications of our dataset include expressive performance analysis and performer identification.
NES-VMDB is a dataset containing 98,940 gameplay videos from 389 NES games, each paired with its original soundtrack in symbolic format (MIDI). NES-VMDB is built upon the Nintendo Entertainment System Music Database (NES-MDB), encompassing 5,278 music pieces from 397 NES games.
XMIDI is a comprehensive, large-scale symbolic music dataset that includes accurate emotion and genre labels, consisting of 108,023 MIDI files. The average duration of the music pieces is approximately 176 seconds, yielding a total dataset length of around 5,278 hours.
YM2413-MDB is an 80s FM video game music dataset with multi-label emotion annotations. It includes 669 audio and MIDI files of music from Sega and MSX PC games in the 80s using YM2413, a programmable sound generator based on FM. The collected game music is arranged with a subset of 15 monophonic instruments and one drum instrument. Source: Image Source: https://arxiv.org/pdf/2211.07131v1.pdf