LISTEN / COMPARE / UNDERSTAND

Hear the difference.
Understand why.

Understanding audio is one thing. Hearing the result is another. Our listening labs use real musical performances, deliberate production choices and carefully explained comparisons to explore what changes the sound.

The ABX Lab brings blind-listening games together in one place: Training, Challenge and Advanced. They are for curiosity, enjoyment and community discussion, not a scientific test of your hearing.

THE RECORDINGS BEHIND OUR LABS

Real music. Real recordings. Real decisions.

We draw on recordings shared for education and research, music from contributing musicians, and material made specifically to test an audio question. Each experiment identifies its actual source. No particular artist, dataset or musical genre defines the whole project.

AlbumDB: a real album opened up for learning

One particularly valuable resource is AlbumDB, developed by Thomas McKenzie, Eloi Moliner and Alec Wright using music from Extraordinary, a ten-song album by musician and audio engineer Tom McKenzie.

Instead of supplying only finished songs, AlbumDB includes the separate recordings at multiple production stages. Researchers and listeners can compare how the same performances change as they are balanced, positioned, processed and mastered.

Explore the four recording stages
01

Raw stems

Individual instrument and vocal recordings without additional mix processing. Some timing edits and removal of silences were already made, so “raw” does not mean completely unedited.

02

Mixed-Dry stems

Human-made level, panning, EQ and compression choices, without time-based effects such as reverb and delay.

03

Fully mixed stems

The processed instrument recordings with the additional effects used in the original stereo production.

04

Finished masters

Completed stereo and immersive 7.1.4 presentations of the songs.

That gives us an unusual opportunity to show what happens between capturing a performance and hearing the finished music. The researchers specifically identify music-production and mixing/mastering education among the dataset's intended uses.

More than one type of music

Our labs can also incorporate original recordings and music contributed by independent artists and collaborators. Recordings from different genres, from intimate arrangements to dense, energetic rock and metal, can help demonstrate how the same techniques behave in different material. Each experiment identifies its source and the permissions governing its use.

How we make comparisons

We aim to hold the performance and starting recording constant when testing a particular decision, explain what we changed, and match listening levels whenever the question requires a fair comparison. Measurements describe levels and signal behaviour. Listening helps us judge how those changes feel and sound. Neither substitutes for the other.

Creative and editorial approval stays with The Sound Question's curator. Our experiments are independently designed adaptations for learning, not replacements for, or proposed improvements to, the original artist's mixes and masters.

CREDITS / RESEARCH

Recording credits and research

Our thanks to the AlbumDB creators for sharing a real album's production stages for study. The original recordings, dataset and educational adaptations are credited below.

Original music: Extraordinary, by Tom McKenzie.
Dataset: AlbumDB: A multitrack dataset of a ten-song album with stereo and immersive 7.1.4 masters (2026).
Research authors: Thomas McKenzie, Eloi Moliner and Alec Wright.
Dataset license: Creative Commons Attribution 4.0 International.

Explore AlbumDB and its source recordings →

Dataset DOI → · License terms → · Research paper (AES Paper 10278) →

AlbumDB-based examples are independently produced educational adaptations by The Sound Question. Their presentation does not imply endorsement, affiliation or collaboration with the artist or research authors. Other recordings will carry their own source credits and permissions.