So here are the results and an explanation on the test.
For each song there is a version that has been provided by the “online mastering” service by BandLab, another by SoundCloud, and another by a human ME (myself). The average loudness (LUFSi) has been matched for each song so this wasn’t playing a role in the decision. There are dynamic range differences, but they are minor.
I ran this test simply because I was curious about what the others would judge as “the best sounding version” compared to my own taste. There is obviously a degree of subjectivity in this test so it isn’t exactly possible to settle on one or the other version as “the best”. However, comparing these versions with commercial hits in similar genres can be a good indicator of a master’s capacity to blend in the industry standard, which I am suggesting you do if you’re curious (and if your opinion changes after this, let me know!).
Without further ado, here are the results.
The figures show the proportion of votes for each version. When there was a tie, one vote has been counted for each version. I ran this test on 4 different forums and private FaceBook groups, gathering roughly 40 different opinions.
A = human master: 35,9%
B = BandLab : 25,6%
C = SoundCloud : 38,5%
A = BandLab : 16,7%
B = human master: 47,2%
C = SoundCloud : 36,1%
It is interesting to note that trends differ from one forum to another. That would possibly tend to demonstrate that there is a bias coming from people reading the answers and being subconsciously influenced with a larger sample but I think the number of votes isn’t large enough to determine that.
So there you have the data. I would like to mention that automated services (not only BandLab and SoundCloud but the numerous options like Landr, emastered etc) are marketed as “online mastering” but are nowhere near resembling how mastering is done professionally for the majority of commercial releases.
These services are only applying automated processing on a 2-bus mix, while human mastering actually entails a lot more:
- Quality control before a song is released forever (typical issues detected by a human: hiss, background noise, clicks & pops, final sustain cut off too soon, editing mishap, etc)
- A fresh view by an experienced human with an artistic sensibility, and the opportunity to apply subtle dynamic enhancements (such as volume automation on a chorus to make it bigger, EQ automation on a specific section with a different arrangement, etc)
- Delivering appropriate file formats for distribution (44,1kHz/16bits for CD, 48kHz+/24bits for digital, MQA or Apple Digital Masters, a cappella or instrumental versions for sync licensing, etc)
- Suitable use of dithering/noise shaping when necessary
- Delivery of DDP images or vinyl premasters
- Metadata embedding (such as ISRC codes, etc)
- The opportunity to have a human feedback on a mix you have obsessed over during weeks…
I hope this was entertaining, or maybe even enlighting.