Every online mastering service has a comparison page, a set of glowing reviews and a demo track that sounds wonderful. None of that tells you how it will treat your song. Reviews date quickly, demo tracks flatter, and the loudest master tends to win simply because it is loudest. This guide gives you a simple, repeatable way to test three AI mastering services (we use LANDR, CloudBounce and BandLab Mastering as the example) on your own mix, with the loudness evened out and the names hidden, so the only thing left to judge is the sound.
You don't need special gear. A pair of headphones you know well, a free loudness (LUFS) meter, any DAW or audio editor, and an hour or so of fresh ears will do. Swap in any three services you like.
Why a Blind, Level-Matched Test Matters
Our ears are easily fooled by volume. Violet Recording's August 2026 guide to A/B comparison puts it bluntly: louder almost always sounds better for the first few seconds. A master that is 1 dB hotter will seem fuller and more exciting even if it has flattened the chorus and pushed the vocal back. The fix is to match the masters' loudness before judging them, measured in LUFS (an average of perceived loudness) rather than by peak level, which says little about how loud something feels.
The second trap is knowing which file is which. If you expect the pricier service to win, you'll hear it winning. Hiding the names removes that bias.
Part One: Prepare One Fair Mix
Step 1 — Finish the mix first. Mastering services can't fix a mix problem, and you don't want to judge them on how they cope with a vocal you were going to rebalance anyway. Make every mix decision now.
Step 2 — Export one clean, unmastered file. Take any limiter or loudness plugin off your master bus and export a single WAV at your project's sample rate, 24-bit if you can, with a few decibels of headroom so the peaks sit comfortably below 0 dBFS (around -6 to -3 dBFS is a common target). Name it something neutral like song-premaster.wav. Every service gets this exact file. If you upload different exports, you're testing your exports, not the services.
Step 3 — Write down what you want. Before you hear anything, jot three or four lines: "vocal clear and forward, kick and bass tight, keep the quiet bridge quiet, loud enough for streaming." This brief stops you being swayed by a master that is impressive but wrong for this song.
Part Two: Get Three Masters on Default Settings
Step 4 — Run the file through each service with comparable settings. Use each service's default or most neutral option for the first round, so you're comparing engines rather than your preset-picking. At the time of writing:
LANDR lets you preview masters free and without limit, and you only pay when you download. It offers several styles that vary in character and loudness, and a reference-mastering option if you want to match an existing track. Downloads range from MP3 to 24-bit HD WAV depending on plan, and a single-track master is sold on its own as well as through subscription tiers.
CloudBounce offers genre-based mastering styles plus post-mastering tweaks, a free preview before paying, and both pay-per-track and subscription plans.
BandLab Mastering is free to use, with presets and an intensity setting. BandLab's help pages list Universal, Fire, Clarity and Tape as the free presets, with others for members; a September 2026 update added finer intensity control to each preset.
Plans and preset lists change often, so check each service's current terms at checkout. For a fair first round, pick the most "balanced" or general-purpose style on each, at a medium intensity, and leave extras like reference matching switched off.
Step 5 — Download the best file each one allows. Ideally a WAV from all three. If one service only gives you an MP3 at the free level, note that down: you'll be comparing a compressed file against lossless ones, which can slightly blur the top end. Better to pay for one full-quality download than to judge on uneven formats.
Part Three: Level-Match and Hide the Names
Step 6 — Measure each master's loudness. Load all three into your DAW or editor and measure the integrated loudness (LUFS) over the same section of each, ideally the whole song or at least the same chorus. Write the numbers down: they show how hard each service pushed.
Step 7 — Turn the louder ones down to match the quietest. Use clip or track gain to bring each master down to the level of the quietest one, so you never add gain and risk clipping. Re-measure until all three are within 0.1 LU of each other; Violet Recording suggests 0.1 to 0.5 dB, and the tighter you get, the fairer the test. Make sure the files are lined up so they start at exactly the same moment.
Step 8 — Blind the files. Ask a friend to rename the three level-matched exports to A.wav, B.wav and C.wav and keep the key on a scrap of paper you don't see. Working alone? Put the three files in a folder, shuffle them with a random-number generator or by drawing lots, rename them, and write the key on a note you fold away. Some referencing plugins have a built-in blind mode that hides which slot is which, which is even easier.
Part Four: Listen and Score
Step 9 — Listen to the parts that matter. Bchillmix's May 2026 comparison suggests checking the first verse, the hook, the bridge and the loudest section rather than listening top to bottom again and again. Switch between A, B and C quickly, because your memory of a sound fades within seconds.
Step 10 — Score each version on four things, out of five. Keep a simple sheet like this:
Version | Loudness feel | Low end | Vocal clarity | Dynamics | NotesA | /5 | /5 | /5 | /5 |B | /5 | /5 | /5 | /5 |C | /5 | /5 | /5 | /5 |
Loudness feel: even matched, does it feel dense and lively, or squashed and tiring?
Low end: are the kick and bass tight and defined, or boomy, thin or smeared?
Vocal clarity: can you hear every word without straining? Bchillmix warns that a heavier setting can make the vocal sound smaller.
Dynamics: does the chorus still lift above the verse, and does the quiet bridge still feel quiet?
Step 11 — Check on more than one system. Score again on earbuds or a phone speaker, and once in the car or on a home speaker if you can. Do a quick mono check too: if the vocal or bass drops away in mono, that's worth knowing before release.
Step 12 — Take a break, then do it again. Come back the next day, have your friend re-shuffle the letters, and score once more. If the same master comes top twice under different letters, you can trust the result. If the winner changes, the differences are small, so let price or features decide.
Step 13 — Reveal the key. Only now do you look at which service was A, B and C. Compare the scores with your brief from Step 3, and with the original loudness readings from Step 6. Often the master you preferred blind isn't the one that measured loudest.
Common Mistakes
Comparing un-matched masters: the loudest nearly always wins, and the test means nothing.
Uploading different exports: use one premaster file for every service.
Leaving a limiter on the mix: it limits what each engine can do and makes them sound more alike.
Mixing file formats: an MP3 against WAVs isn't a fair fight.
Peeking at the key: once you know, you'll hear what you expect.
Judging on one playback system: what sounds great in headphones can fall apart on a phone.
Trusting old reviews: engines and plans change, which is the whole reason to test for yourself.
A Quick Checklist
1. Mix finished, no limiter on the master bus2. One 24-bit WAV premaster with headroom3. Brief written before listening4. Same file through all three, neutral settings5. Best-quality download from each, formats noted6. Loudness measured, matched within 0.1 LU7. Files renamed A/B/C, key hidden8. Scored on loudness feel, low end, vocal, dynamics9. Re-checked on phone, car or speaker, plus mono10. Repeated next day with a fresh shuffle
Where That Leaves Us
A good master is one that serves the song, not one that wins a volume contest. An hour of blind, level-matched listening on your own music tells you more than a dozen reviews, and you can repeat it whenever a service updates its engine. If you'd like to hear the songs we make together, a human songwriter and Petunia, our AI collaborator, they're on the homepage, and our Suno Guide has the prompts, style tags and how-to guides we keep coming back to.
Sources
Violet Recording — How to A/B your mix (louder sounds better, matching by LUFS within 0.1–0.5 dB, blind switching, quick switching, checking on several systems and in mono; August 2026)
Bchillmix — Best AI mastering services compared in 2026 (same clean WAV for every service, level-matched comparison, which sections to check, louder presets shrinking the vocal, CloudBounce genre styles; May 2026)
LANDR — Mastering (unlimited free previews, styles varying in intensity and loudness, reference mastering, accepted and download formats, single-track purchase)
BandLab Help — How do I master my music on BandLab? (free and member presets, intensity settings, the mastering site)
Magnetic Magazine — BandLab Mastering update gives artists more control before release (finer intensity control per preset; September 2026)
Violet Recording — CloudBounce review (free preview before paying, pay-per-track and subscription options; August 2026)
