How to master a song for free, right in your browser.
Build and audition a mastering chain in your browser: load a finished mix, use the heuristic assistant as a starting point, compare carefully, then export locally.
What mastering actually does
Mastering is the last creative step before a song goes out. Your mix already has every instrument balanced; mastering treats that stereo file as one finished object and shapes its overall tone, density and loudness so it translates well on phones, laptops, car stereos and big speakers — and so it sits at a competitive volume next to other releases on Spotify, Apple Music or YouTube.
In a traditional studio that means a chain of EQ, compression, saturation, stereo work and a limiter. The good news: you don't need that gear, a paid plugin bundle, or an expensive engineer to get a solid result. Learning how to master a song for free is mostly about understanding the order of operations and listening carefully — and AudioWrench's Mastering Studio packages the whole chain into one page.
How to master a song for free, step by step
Here's the real flow inside the Mastering Studio, from a raw mixdown to an exported master:
- Accept the current mix, not the previous one. Choose a supported WAV, MP3, FLAC or other file of up to 200 MB, or try the demo. Checking, decoding and estimating loudness take time and depend on device memory. Wait for the current-track status. Replacement immediately clears old audition, exports and frozen measurements; failed or cancelled loads do not restore old audio.
- Run the Master Assistant. Pick a target — Pop / Streaming, Hip-Hop / Trap, EDM / Club, Rock / Metal, Acoustic / Jazz, Lo-fi / Chill or Podcast / Voice — then hit Analyze & build my chain. It reads your loudness, low end and brightness and dials in the whole chain for that genre. Prefer to match a song you love? Drop it into the reference track box and the EQ moves gently toward its tonal balance.
- Fine-tune one decision at a time. The rail shows EQ, Bass Control, Inflator, Exciter, Stereo Width, Glue and Maximizer. Toggle a module to bypass its processing without losing values. Drag to reorder, or focus its tab and use Alt + Left/Right. If a pending Assistant or reference match is still running, a manual chain edit cancels that older result instead of letting it overwrite the edit.
- Preview as you go. Use the seekable, loopable transport to play, A/B against the source, and loop a chorus while you adjust. You can even hit Preview codec to hear how the master will sound after MP3 encoding before you commit.
- Set a loudness reference. The Maximizer uses Web Audio dynamics processing while the studio estimates source and rendered LUFS plus sample peak. It is not certified true-peak or brick-wall limiting, so leave headroom and verify important deliverables separately.
- Render, request and independently check the file. Set IN/OUT to the intended delivery range, then render. A source, range or chain edit makes that render obsolete. Export WAV 16-bit with optional TPDF dither, WAV 32-bit float or MP3 320. The current receipt shows the generated filename, format, PCM frame count, sample rate and size. Reopen the actual saved file: a browser download request is not proof of successful saving, distributor acceptance or compliant true peak.
What the seven module displays are actually telling you
Settings preview is a calculated response to control values, not a measurement of your mix. Live measured uses input/output taps during Play. Freeze holds an earlier measurement until unfrozen or replaced by a new source; editing a setting does not turn a frozen measurement into a new one. Pre/Post hides a trace, whereas the module's enabled toggle changes processing.
| Module | Useful visual cue | Decision to verify by listening |
|---|---|---|
| Master EQ | Six-band response plus measured pre/post spectra; drag the response nodes or use the sliders. | Does a small bass cut or presence change help at matched loudness? A silent band cannot supply a measured delta. |
| Bass Control | Low-frequency spectral comparison and actual compressor gain reduction when enabled. | Is the low end controlled without losing kick impact or making bass notes uneven? |
| Inflator | Transfer-curve preview, input/output level change and measured sample clipping. | Does extra density add useful weight, or make transients and sustained tones rough? |
| Exciter | Mode response and measured high-frequency pre/post balance. | Is useful detail clearer without bringing out hiss or harsh cymbals? |
| Stereo Width | Stereo/phase view, correlation and mid/side information. | Does widening improve space without a weak center? Correlation is a warning cue, not a guarantee of mono compatibility. |
| Glue | Threshold/ratio transfer curve and measured gain-reduction history. | Do hits recover naturally with your attack/release choices, or does the mix pump? |
| Maximizer | Push/ceiling settings and processed-path gain-reduction history. | Are attacks, bass and quiet detail still useful? Sample-peak readings do not certify inter-sample true peaks. |
The top spectrum follows the audition you selected, including Source or Difference. The Maximizer reduction meter follows the processed chain even while you audition Source. The source → render LUFS field describes the last accepted render, not a continuously updated loudness measurement of every slider move.
Audition loop, master range and level match are separate
A loop is a repeatable listening passage. IN/OUT defines the downloaded range; the render keeps that range's frame count rather than adding the audition loop or a reverb tail. Use Full track before exporting the whole song if you previously trimmed a passage. Level match only attenuates the louder audition branch and does not apply an extra boost to the final file. Difference is Mastered minus synchronized Source at the chosen comparison levels, not a separate exported mix.
Worked example: check a dense chorus without exporting just the chorus
Keep your untouched mix elsewhere. Load it, use an Assistant target as a starting chain, then loop the busiest chorus. If the bass masks the kick, compare Bass Control Pre/Post while making a small correction. Check Glue reduction on hits and undo settings that flatten attacks. Listen to the quiet passage too: one loud chorus cannot tell you how the whole song behaves.
Return to Source with level match enabled. If the change only seems better when louder, reconsider it. Select Full track, confirm both time fields, render, then try a lower-bitrate codec audition for fragile high-frequency detail. Stop codec audition before normal Play. Export your chosen format and reopen the saved file to confirm the whole song, channel count, boundaries and sound.
Cancel, replace and recover without accepting an older master
- New track pending: Cancel load or Escape discards that selection. Previous audio and results stay retired. Choose the previous file explicitly if you want to return to it.
- Analysis or pre-stage pending: Cancel pending work keeps the accepted source and editable settings but rejects the unfinished result. A newer manual edit also retires older Assistant/reference work.
- Render, encoding or codec pending: cancellation rejects the output and any later playback/download. The browser may still need to finish an unabortable task; wait for processing to settle before retrying instead of launching duplicate renders.
- Decode or render fails: follow the inline message, try another supported source or retry the current render. Rendering errors keep the accepted source/settings; failed replacement loads leave no active source.
- Leaving the page: pending work and playback are retired. Returning is not permission for an older codec to restart automatically.
Choose an export and verify delivery separately
WAV 16-bit exports integer PCM; its TPDF dither option is for this conversion, not a loudness effect. WAV 32-bit float retains a floating-point PCM delivery format. MP3 320 is a lossy encoded copy; preview bitrate does not change this fixed export bitrate. All use the current rendered IN/OUT range. Choose the recipient's required format rather than assuming every distributor accepts every option.
Keep the original until you have reopened the delivered file. Check the complete intended passage, starts/ends, channel balance and any encoding artifacts. Use an appropriate compliant meter for critical true-peak and delivery checks. The local receipt describes generated bytes and source settings; it does not inspect the recipient, your operating-system Downloads folder or a platform's later transcoding.
Getting a clean, loud master
The most common mistake is chasing loudness with the limiter alone. Push too hard and the maximizer squashes your transients — kicks lose punch, the whole track sounds flat and fatiguing. The fix is to do the loudness work earlier in the chain: tame harsh resonances with EQ, control the low end so the bass isn't eating headroom, add a touch of glue compression for cohesion, and only then let the maximizer bring it up to target. If a previous master already sounds over-squashed, the optional Transient Revival pre-stage can bring the hits back out.
Trust your ears over the meters, but use the meters as a sanity check. Master at a comfortable volume, take breaks so your ears stay honest, and compare against a commercial reference in the same genre. Headphones are strongly recommended for catching stereo and low-end detail.
🎚️ Try the Mastering Studio (free) →
Why master in your browser?
The Mastering Studio decodes and processes the selected file on your own device through the Web Audio API; the audio payload is not uploaded to AudioWrench. Normal page assets still load over the network. That matters for unreleased work because the processing path does not require a cloud-file handoff. There's no account to create, plugin to install, subscription or watermark, and the downloadable render is produced locally from the selected settings.
FAQ
Can I really master a song for free?
Does my song get uploaded to a server?
How loud should my master be?
Do I still need to mix before I master?
Sources and review notes
Loudness targets are playback and delivery context, not a universal mastering recipe. AudioWrench measurements are useful estimates and should be paired with listening references.
- ITU-R BS.1770-5 — Loudness and true-peak measurement algorithms.
- Spotify loudness normalization — Spotify’s current playback-normalization behavior and delivery guidance.
- AudioWrench review methodology — Scope and limits of AudioWrench implementation claims.