How Does AI Mastering Work? A Plain-English Guide
How AI mastering analyzes and processes your track, EQ, loudness, and true peak explained in plain English with real LUFS numbers.
You can master to a reference track online for free by uploading (or, in TrackGleam, loading in your browser) a song you want to match, giving it a reference track you love, and letting the tool copy that reference's tonal balance and loudness onto your song. TrackGleam does this fully in the browser — no Python install, no weekly song cap — and level-matches the A/B so you judge tone, not volume.
Reference-track mastering means shaping your song so its overall tone, stereo width and loudness resemble a professionally released track you admire — your "reference." Instead of guessing EQ moves in the dark, you point the tool at a song whose sound you want, and it measures the frequency balance, peak level and loudness of that reference, then nudges your master toward the same profile.
It is the most concrete way to answer the eternal question "why doesn't my mix sound like a real record?" The reference becomes an objective target. This is a well-established mastering technique; the open-source project Matchering describes it as matching your target track's RMS, frequency response, peak amplitude and stereo width to the reference's. If you want the broader picture of what the underlying engine is doing, see how AI mastering works.
Every reference matcher follows the same three-step logic. First, it analyzes the reference track: what its frequency curve looks like, how loud it is, how wide the stereo image sits. Second, it analyzes your song the same way. Third, it applies corrective EQ, gain and (sometimes) width and dynamics processing so your song's measurements move toward the reference's.
The catch worth understanding: matching tone is not the same as cloning a song. A reference matcher copies the spectral balance and loudness, not the arrangement, the instruments or the vibe. A sparse acoustic ballad matched to a wall-of-sound pop record will get brighter and louder, but it will not gain the missing layers. Pick a reference in the same genre and instrumentation as your song for the most believable result.
The three common ways to do reference mastering without a paid studio each have a real trade-off. Matchering is genuinely excellent and free, but it is an open-source Python library — you either install it and its dependencies yourself, or self-host the included web app. Songmastr is a hosted service built on similar ideas with a friendly web page, but its free tier is capped. TrackGleam runs the match in your browser with no install and no weekly limit.
| Option | Install needed? | Free-tier limit | Level-matched A/B? |
|---|---|---|---|
| Matchering (open source) | Yes — Python or self-host | Unlimited (you run it) | Not built in |
| Songmastr | No (web) | 7 songs per week free, per their site | Not stated |
| TrackGleam | No — runs in the browser | Free reference match, no weekly cap | Yes, auto level-matched |
Verified July 2026 — prices/specs change; re-check the source.
Details as of July 2026: Songmastr's own site advertises 7 free masters per week and notes a 10-minute / 80MB per-file limit (songmastr.com/faq). Matchering's capabilities and open-source status come straight from its GitHub repository. If you are comfortable with a terminal and want unlimited free processing, Matchering is a superb choice — TrackGleam's pitch is simply "same idea, zero setup, and honest metering."
Here is the trap that ruins most reference comparisons: louder almost always sounds "better" for the first few seconds. If your matched master comes back louder than the original and you A/B them at the same fader, your ears will prefer the louder one regardless of whether the tone actually improved. That is a loudness bias, not a quality judgment.
A level-matched A/B automatically trims the two versions to the same perceived loudness before you compare, so the only difference you hear is the tonal and dynamic change — the thing you actually care about. This is the same principle behind making a whole playlist feel even; if you are chasing consistency across songs, our guide to making all your songs the same volume covers it in depth. Without level matching, "it sounds better" often just means "it got louder."
The workflow in TrackGleam takes a couple of minutes and never uploads your audio — the processing happens locally in your browser via Web Audio and WebAssembly:
If your reference is a loud modern record and you plan to release to Spotify or Apple Music, remember that platforms normalize playback — chasing a crushed reference can backfire. Our LUFS streaming targets guide explains where to land so your master survives normalization intact.
The honest test of any reference matcher is whether the result actually sounds like the target — and you should never have to pay to find that out. With TrackGleam you preview the full match in the browser for free, level-matched against your original, before any money changes hands. If it nails the tone you wanted, great; if the reference was a poor fit, you have lost nothing but two minutes.
Paid AI masters start at $1.99 for a single track (or $99 lifetime with a fair-use monthly cap), with a 14-day money-back guarantee — but the reference-match preview and the free -14 LUFS master cost nothing and require no account. Many "free" mastering tiers around the web are preview-only or bury a watermark or a weekly cap in the fine print; the fix is simply to check what you actually get before you commit, which is exactly what a real, full, level-matched preview lets you do.
Yes. TrackGleam lets you load your song and a reference track in the browser and hear the level-matched match for free before paying. The open-source tool Matchering is also free if you install it yourself, and Songmastr offers a limited free tier (7 songs per week as of July 2026).
For most people, yes. Matchering is powerful and free but requires installing Python or self-hosting the web app. TrackGleam does browser-based reference matching with no install and no weekly cap, plus an auto level-matched A/B and verifiable LUFS/true-peak metering. If you are comfortable with a terminal and want unlimited local processing, Matchering remains an excellent choice.
No. Reference matching copies the tonal balance, loudness and stereo width of the reference — not its arrangement, instruments or performance. A sparse mix matched to a dense pop record gets brighter and louder but does not gain missing layers. Choose a reference in the same genre and instrumentation for the most believable result.
Because louder almost always sounds better for a few seconds, an un-matched comparison tricks your ears into preferring the louder version regardless of tonal quality. A level-matched A/B trims both versions to the same loudness so you judge only the tonal and dynamic change, which is what actually matters.
No. TrackGleam processes audio entirely in your browser using Web Audio and WebAssembly. Your song and your reference track never leave your device — there is no upload, no account and no login required to try it.
TrackGleam works with WAV and MP3. The free master targets roughly -14 LUFS integrated with a -1.0 dBTP true-peak ceiling for safe streaming delivery, and it measures integrated LUFS, true peak and loudness range on the finished file so you can verify the result in any meter.
function sub() { [native code] }
How AI mastering analyzes and processes your track, EQ, loudness, and true peak explained in plain English with real LUFS numbers.
Make all your songs the same volume free in your browser — no uploads, no account. Level one track free, or batch a whole folder to a verified −14 LUFS target.
Every platform's LUFS and true-peak target in one dated table — Spotify, Apple Music, YouTube, Amazon, Tidal, TikTok — and how to hit -14 LUFS free.
LANDR vs eMastered on price, sound, cloud uploads and subscription lock-in, plus a no-subscription browser option that undercuts both.
Every guide and comparison, in one place.