-- Presented on ISMIR 2022! --


8970 five-second length audio recordings of real electric guitar sounds

All notes of a standard tuning electric guitar processed through 12 different effects


EGFxSet contains recordings of all clean tones in a Stratocaster guitar, with augmentations by processing the audios through twelve electric guitar effects. Similar datasets apply effects using software, EGFxSet in contrast uses real guitar effects hardware, making it relevant to develop MIR tools with applications on real music. Annotations include all guitar and effect parameters controlled during our dataset recording. EGFxSet contains 8970 unique, annotated guitar tones, and is published with full open-access rights.


All possible notes of an standard tuning electric guitar were recorded using the instrument input of an Audient iD14 audio interface. Using the interface output, the recordings of each sound effect were processed for recording using the line input of the same interface. Both, the previously recorded clean guitar sounds and the processed recordings were normalized.

All the sounds of the dataset were annotated by a professional guitarist. Each file’s annotations are:

  • String-Fret tuple: digits indicating the string number (1 through 6) and fret number (0 for open strings, and 1 through 22 for fret position).

  • Note: the guitar had standard tuning, so the note played can be inferred with string-fret tuple.

  • Pickup configuration: one of the five possible pickup configurations: bridge, bridge-middle, middle, middle-neck, or neck.

  • Effect: this is the name of the effect used (i.e. blues driver, chorus, or plate reverb).

  • Model: this is the commercial name and version of the ef120 fect hardware used.

  • Effect type: a categorical name for the effect type, abstracting away the effect’s specific name and brand.

  • Knob names: each effect we used has controllable parameters with names given by the original manufacturer.

  • Knob type: we use language descriptors from our functional categorization of effect controls to categorize each effect’s controllable parameters.

  • Setting: a value (between 0.0 and 1.0, or bpm for rate effects, or name of selected category) indicating the setting captured.


Existing datasets use software emulations of analog gear to apply effects on clean guitar recordings, and few include effect parameter annotations. As a result, there is a need for datasets featuring recordings of clean electric guitar tones, augmented using real effects hardware, and including annotations of effects and guitar settings.

EGFxSet addresses this need!

Content type performance performance mono- and poly-phonic mono- and poly-phonic monophonic
No. guitars 1 1 2 2 1
No. performers 6 1 1 1 1
No. effects 0 5 11 13 12
No. guitar pickup settings 1 1 2 2 5
Effects with real hardware - False False False TRUE
Annotated effect settings performance performance mono and polyphonic mono and polyphonic monophonic
Duration (clean) 3 h 2 h 35 min 34 min 57 min
Duration (effects) - 10 h 18 h 33 min 111 h 6 min 11 h 30 min
Duration (total) 3 h 12 h 19 h 8 min 111 h 40 min 12 h 27 min

Overview of existing datasets featuring (electric) guitar recordings and augmentations with audio effects. Each row contrasts their characteristics with EGFxSet. Note that EGFxSet is the only one using real effects hardware.

Content type monophonic
No. guitars 1
No. performers 1
No. effects 12
No. guitar pickup settings 5
Effects with real hardware TRUE
Annotated effect settings monophonic
Duration (clean) 57 min
Duration (effects) 11 h 30 min
Duration (total) 12 h 27 min

Overview of the characteristics of EGFxSet.


American Strat

All 138 possibles notes with each of the 5 pickups of an American Fender Stratocaster plucked by a 2.0 mm thick pick

Orange Crush 60

Carefull annotations of each note, string and fret position, and the effect parameters applied

American Strat
Orange Crush 60
American Strat
Orange Crush 60
American Strat
Orange Crush 60
American Strat
Orange Crush 60

Analog gear effects with and without digital components
12 popular effects are divided into four categories: Gain, Modulation, Delay, and Reverb
3 different effects per category


TubeScreamerMini: A compact version of the iconic TS-808 Saturation effect from Ibanez. The pedal used was made in Japan in 2019.

BD-2: A widely used overdrive pedal from Boss Company. The version used is a 2018 model made in Taiwan and has the same schematic as their first Blues Driver pedal released in 1995.

RAT2: A Proco’s very popular distortion pedal, the one used was a later 2010 model made in China by Neutrik.


CE-3: A Chorus Pedal made in Japan in 1987 by Boss. It includes, at its time, a stereo configuration, but we only used the mono one.

Phase45: A MXR phaser pedal made from 1977 to 1981, its a simplified one knob version of their well known phaser pedal Phase90, the one used was build in U.S.A. in 1980.

E-Lady: A Mooer clone of the Electric Mistress, an iconic flanger pedal from 1975 made by Electro-Harmonix. The pedal used was made in China in 2022.


DL4: A delay stomp box unique in its kind when introduced into the market in 1999. This digital emulator of 16 different type of delays remained unmodified for nearly 23 years. Of the 16 possible delays we used three, Tape Echo, Sweep Echo and Digital Delay. The pedal used is a Line 6 2007 model made in China.


CR6OC: We used a digital reverb emulations of an Orange Crush Pro 60 12-inch speaker combo amplifier. The amplifier used is a 2014 model made in China and it integrates Plate, Hall and Spring reverberations.

Effects Settings

Blues Driver

(Boss BD-2)

Tube Screamer

(Ibanez TS Mini)


(Pro Co Rat2)


(Boss CE-3)


(Mooer E-Lady)


(MXR Phase45)

Tape Echo

(Line6 DL-4)

Digital Delay

(Line6 DL-4)

Sweep Echo

(Line6 DL-4)

Plate Reverb

(Orange CR-60 Combo Amp)

Hall Reverb

(Orange CR-60 Combo Amp)

Spring Reverb

(Orange CR-60 Combo Amp)

The table illustrates the guitar effect hardware settings used when recording EGFxSet. Feel free to click in each cell to hear a sample of the sound of each effect!


The current release of the dataset can be downloaded via its zenodo repositorie or used via its mirdata loader. The dataset structure and steps needed to use its files are carefully specified in the links below.

Link to Zenodo

EGFxSet Zenodo repositorie

Link to Mirdata

EGFxSet Mirdata loader


A multidisciplinary team with a strong music foundation composed of researchers from two different universities
Irán Roman
Postdoctoral Researcher of New York University (NYU)
Hegel Pedroza
M.A. Student of National Autonomous University of Mexico (UNAM)
Gerardo Meza
M.A. Student of National Autonomous University of Mexico (UNAM)

For any inqueries or futher details, feel free to contact any author through their personal email or social media accounts or send an email to: iran@ccrma.stanford.edu