Should You Upload WAV or MP3 to an Online Mastering Service?
Upload WAV, not MP3. Every reputable online mastering service wants a 24-bit, 44.1 or 48 kHz WAV file with at least 3-6 dB of headroom, no master limiter engaged, and no dithering. Uploading MP3 caps the ceiling on what the service can do because the high-frequency content and transient detail you need for loud, competitive mastering has already been thrown away by the compression codec.
The file format decision is made before you open a mastering account. If the export you send is wrong, the master comes back wrong — and you cannot fix a compromised source in the mastering chain after delivery.
If you want a release-ready master that survives streaming, bouncing a clean 24-bit WAV with headroom is the first step that actually matters.
Book Mastering ServicesWhat Actually Happens to an MP3 During Mastering
MP3 is a lossy format. The encoder throws away frequency information it judges to be inaudible — the upper treble above 16 kHz on standard bitrates, subtle transient detail on percussive elements, and stereo information the codec flagged as redundant. That data is gone. No mastering tool on the planet rebuilds it.
When a mastering engineer or an AI chain pushes that MP3 through an EQ, a multiband compressor, and a limiter, the compression artifacts get amplified alongside the music. The "swirl" in the high-hat, the pre-echo on the kick, and the grainy character on sibilants all get louder. A good master on a bad source still sounds like a polished version of a bad source.
The other problem is headroom. MP3 files often peak at or near 0 dBFS because they were bounced from a finished mix with a limiter on it. You cannot run a full mastering chain on a file that is already clipped.
The Correct Export Settings for Online Mastering
Every major online service — whether human-engineered or AI — has near-identical upload specs. Hit these and you eliminate 80 percent of the "this mix came back weird" complaints:
| Setting | Target | Why it matters |
|---|---|---|
| File format | WAV or AIFF | Lossless; preserves everything the mastering chain needs |
| Bit depth | 24-bit | Avoids digital-low-level noise when the service brings up quiet sections |
| Sample rate | Match your session (44.1 or 48 kHz) | No sample-rate conversion artifacts from a resample before upload |
| Peak level | -3 to -6 dBFS true peak | Gives the limiter room to breathe, keeps transients intact |
| Master bus limiter | Off | The mastering engineer's limiter replaces yours, not layers on top |
| Dithering | Off (at 24-bit) | Only dither at the final 16-bit step, not before mastering |
| Normalizing | Off | Normalizing erases the dynamic shape the engineer needs to see |
If your DAW lets you, use a true-peak meter (not a sample-peak meter) when checking your export. A file that reads -3 dBFS sample peak can still read -0.5 dBTP after ISP reconstruction, and that half-dB of hidden inter-sample clipping is what makes a master sound harsh on phones.
Why Some People Still Upload MP3 (and Regret It)
Three reasons come up over and over:
- Storage and upload speed. A 3-minute WAV is roughly 30-50 MB. A 320 kbps MP3 is 7 MB. On a slow connection this feels like a meaningful difference — but it is the wrong place to save time.
- Loudness habit. Producers who upload a limited, normalized MP3 because "that is how the song sounds" do not realize the final loudness belongs to the mastering stage, not the mix stage.
- Lost project file. If the only copy of the song is a bounced MP3, uploading that is sometimes the only option. In that case the master will still sound better than nothing — but call it a rescue, not a real master.
None of these justify uploading MP3 on a song you care about. A song worth paying to master is a song worth re-bouncing correctly.
How to Bounce a Mastering-Ready File in Popular DAWs
The language is slightly different in every DAW, but the steps are the same. Remove the final limiter from the master bus. Pull the master fader down until peaks sit between -3 and -6 dBFS. Set export to WAV, 24-bit, at your session sample rate. Turn dithering and normalizing off. Render.
If you work in FL Studio, the guide to exporting vocal stems from a FL Studio template walks through clean file preparation in detail, including why labeling and routing matter before rendering. The same logic applies if you are mixing in Ableton, Logic, Pro Tools, or Reaper: disable the limiter, leave headroom, bounce WAV.
WAV vs AIFF vs FLAC
Most services accept all three. Practical differences:
- WAV: the universal default. Every DAW and every mastering platform handles it without drama. If you are unsure, send WAV.
- AIFF: identical audio quality to WAV, but with better metadata support on Apple-based workflows. No sonic advantage, no disadvantage.
- FLAC: lossless compression — smaller file size, identical decoded audio. Fine to send if the service explicitly supports it, but many AI platforms still reject FLAC. WAV is safer.
Never send a MIDI export, a lossy format like AAC or OGG, or a mobile-converted file from a voice-memo app. Those go through the same quality-degrading encoders as MP3.
The Loudness Target Question
If you are preparing for streaming, the loudness target is outside your export. Spotify normalizes to around -14 LUFS integrated, Apple Music to around -16 LUFS, YouTube to roughly -14. Your mastering engineer or mastering service will hit the loudness that fits the release. You should not try to pre-master the file to -14 LUFS before uploading. Send a clean, unlimited, -3 to -6 dBFS peak WAV and let the mastering stage do its job.
For a deeper look at how to prep a mix before it reaches mastering, read what to send a mastering engineer before you order a master. It covers the technical checks that reduce revision rounds.
When AI Mastering Forgives a Bad Source (and When It Does Not)
AI mastering platforms have gotten very good at flattering a mediocre mix. What they cannot do is un-compress an MP3. If you upload a 128 kbps MP3 to an AI service, it will return a louder version of a 128 kbps MP3. The algorithm has no way to invent the frequency detail the codec discarded.
A human mastering engineer can at least decide whether a lossy source is worth spending time on, and communicate expectations. An AI service will simply process whatever you send. That makes the format question more important for AI services, not less.
What to Do If You Only Have an MP3
If the only surviving file is an MP3, be honest about the job. You are not doing ideal mastering; you are doing restoration and rescue. Send the highest-bitrate MP3 you have, avoid converting it to WAV and pretending it was always lossless, and tell the engineer where the file came from. Converting MP3 to WAV does not restore quality. It only places the same lossy audio inside a WAV container.
A rescue master can still be useful for an old song, a social release, a catalog cleanup, or a reference upload. The engineer may reduce harshness, rebalance broad tone, control peaks, and make the song sit closer to your newer releases. The limits show up when the source has swishy cymbals, grainy vocal esses, smeared transients, or clipping from the original bounce. Those problems can be softened, but they cannot be fully undone.
If you can reopen the DAW session, do that instead. Even if the mix is old, a clean 24-bit WAV bounce from the original session gives the mastering chain far more to work with. Remove the limiter, leave headroom, export the mix at the original sample rate, and label the version clearly. That one extra step is often more valuable than any plugin used in mastering.
How to Check the File Before Uploading
Before paying for mastering, open the exported file and inspect it like a deliverable. First, listen from beginning to end for clipped sections, accidental silence, missing intro tails, cut-off reverb, or wrong versions. Then check the loudest point. The mix should not already be pinned to 0 dBFS. If it is, go back to the session and lower the master output or remove the limiter.
Next, confirm the file name. A good file name might be `Artist_SongTitle_Mix03_24bit_48k.wav`. That tells the engineer or platform what the file is without opening it. Avoid names like `finalfinal2.mp3`, `bounce.wav`, or `master this one maybe.wav`. File naming sounds minor, but when several versions exist, unclear naming is how the wrong mix gets mastered.
| Check | Good Upload | Problem Upload |
|---|---|---|
| Format | 24-bit WAV or AIFF | MP3, AAC, screen-recorded audio |
| Limiter | Final limiter off | Limiter printed for loudness |
| Peaks | Headroom remains | Peaks clipped or pinned |
| Version | Clear mix number | Unclear "final" naming |
| Source | Original DAW bounce | Converted file from a lossy source |
Why WAV Is Still the Safest Default in 2026
Streaming platforms and mastering tools have changed, but the safest upload format has not. WAV is still the simplest answer because it is uncompressed, widely supported, and easy for both human engineers and automated systems to ingest without guessing. AI mastering platforms may accept several formats, but acceptance is not the same as best practice. If a service accepts MP3, that only means it can process the file. It does not mean MP3 is the right file to send.
WAV also protects the decision chain. If the master comes back too bright, too loud, or not loud enough, you know the problem is in the mastering decision or the mix itself, not in a lossy upload. With MP3, every artifact becomes another variable. The engineer has to work around codec damage before judging the real mix. That slows down revisions and makes feedback less precise.
For independent artists, the rule is simple: archive the clean mix WAV, send that WAV to mastering, then create compressed formats only after the master is approved. MP3 belongs at the distribution or sharing stage, not at the mastering input stage.
How File Format Changes the Mastering Decisions
Mastering is full of small decisions. The engineer may add a little air, tighten the low end, control harshness, widen the image, or push the limiter. Those decisions assume the source file still contains clean transients and full-bandwidth information. A WAV file gives the chain room to make those choices. An MP3 source forces the chain to treat artifacts like part of the music.
For example, a bright vocal on a WAV may need a small de-esser move or a careful high-shelf adjustment. A bright vocal on an MP3 may have codec grain around the same sibilant area. If the engineer reduces that grain too much, the vocal gets dull. If the engineer leaves it alone, the master gets harsh. That is the kind of tradeoff that happens when the upload file is already damaged.
Low end has a similar issue. A good WAV export preserves the shape of the kick and 808. A low-bitrate MP3 can blur the transient and make the limiter react less predictably. The final master may still get loud, but it may feel flatter, less punchy, or less stable on small speakers.
Common Export Mistakes That Look Like Format Problems
Sometimes the issue is not WAV vs MP3. It is the way the WAV was exported. A clipped WAV is still a bad source. A normalized WAV with the master limiter printed is still a bad source. A WAV bounced from the wrong mix version is still the wrong file. Lossless format protects quality, but it does not fix careless exporting.
Before uploading, make sure the master bus has no final loudness limiter, no unnecessary clipping, and no "make it loud" processing that was only there for demo playback. If you love the sound of a little bus compression, you can leave gentle glue in place. If the processor is mainly there to win a loudness comparison, turn it off and let mastering handle that stage.
Also check the start and end of the file. Do not cut the first transient. Do not trim off the reverb tail. Do not export a loop selection by accident. Online mastering services process exactly what you send. If the tail is gone, the master will not grow it back.
What to Send Along With the WAV
The WAV is the main file, but context helps. Send the song title, artist name, mix version, BPM, key if known, and one or two reference tracks. If the master is for streaming, say that. If the master is for a video, say that. If you need a clean version, instrumental, or performance track, mention it before mastering starts.
If you are using a human mastering service, explain what you like about the rough mix and what worries you. "The vocal level feels right, but the 808 may be too soft" is useful. "Make it professional" is not. Mastering feedback should focus on translation, loudness, tone, low end, top end, and overall finish. It should not introduce mix changes that require stem access unless you are ordering stem mastering.
If you are using AI mastering, upload the best WAV and choose the closest style setting. Do not use AI mastering to repair an unfinished mix. AI can shape tone and loudness, but it cannot make a vocal sit correctly if the vocal is buried in the stereo file. The cleaner the WAV, the better the automated result.
The Simple Rule for Release-Ready Masters
If the song matters, send WAV. If the only file is MP3, call it a rescue. If you still have the DAW session, rebounce the WAV. If you are unsure which version is right, send the clean mix export with no limiter and ask the mastering engineer before paying. That simple habit prevents most file-format mistakes.
Mastering should be the final polish on a strong mix, not a repair attempt on a compromised upload. The easiest way to give the master a fair chance is to send the file format that preserves the mix: a clean, lossless WAV or AIFF with headroom.
That also gives you a cleaner archive. Years later, if you need an instrumental master, a remaster, a sync version, or a new loudness target, the original WAV mix is still useful. A mastered MP3 is not a long-term production asset. Keep the lossless mix, the approved master, and any instrumental or clean versions organized in the same release folder.
That archive habit also protects collaboration. If a label, playlist team, video editor, or mastering engineer asks for a clean source, you can send it immediately instead of rebuilding the session under pressure. Good file discipline feels boring until the day it saves a release.
Quick Decision Checklist
- Is the master bus limiter off? If not, remove it before bouncing.
- Does the loudest peak read between -3 and -6 dBFS? If not, pull the master fader.
- Is the export set to 24-bit WAV at your session sample rate? If not, change it.
- Is dithering off? If you are exporting 24-bit for mastering, leave it off.
- Does the file name include the song title, the mix version, and the BPM? That saves a revision round later.
If you hit all five, you are sending a mastering-ready file. The decision between WAV and MP3 is never really a toss-up — it is the difference between giving the mastering chain something to work with and hoping it hides damage you already did.
FAQ
Can a good mastering engineer save an MP3 source?
They can make it louder and more cohesive, but they cannot restore lost detail. A good engineer will usually tell you the limits of what is possible with an MP3 source and either give you a rescue master or ask for a WAV re-bounce. If you care about the release, send WAV.
Does 32-bit float help over 24-bit WAV for mastering upload?
Only if the receiving service explicitly supports 32-bit float, which many AI platforms still do not. 24-bit is the safe universal standard and sounds identical once the file is inside the mastering chain.
What sample rate should I send?
Send whatever your session ran at — 44.1 kHz is fine for most hip-hop and pop, 48 kHz is common for anything crossing into video. Avoid up-sampling or down-sampling before export. Any sample-rate conversion should happen inside the mastering chain or at the final delivery stage, not twice.
Should I send stems or a 2-track WAV for mastering?
Default is a 2-track stereo WAV. Stems are for stem mastering, which is a different service tier and usually costs more. If you are unsure whether your mix needs stem mastering, look at whether the mix has a specific problem a mastering engineer cannot fix in stereo — like a vocal that is too loud against the beat. If yes, consider stems. If not, 2-track is the standard deliverable.
Is it okay to leave a little compression on the master bus before mastering?
Light bus compression for glue is fine, as long as it is not crushing the mix and there is still 3-6 dB of peak headroom. What must come off is the final brick-wall limiter. Anything that defines the final loudness belongs to mastering.
Should I convert an MP3 to WAV before uploading it?
No. Converting MP3 to WAV does not restore the audio that MP3 compression removed. If the original WAV is gone, tell the mastering engineer the source is MP3 and send the highest-quality version you have. If the DAW session still exists, re-export a real WAV instead.
