Mastering for Streaming: Spotify and Other Destinations
The purpose of streaming mastering is to deliver a musically convincing, technically sound recording. Loudness recommendations are playback engineering guidance, not rules for classifying a source or hiding AI provenance.
Spotify's published reference values
Source checked: October 2026. These are Spotify's own published recommendations, not measurement guarantees for every device.
| Parameter | Published guidance | Engineering interpretation |
|---|---|---|
| Normal playback loudness | −14 LUFS integrated (ITU-R BS.1770) | A normalization reference applied at playback; not a universal creative loudness requirement |
| True peak on the recommended target | Below −1 dBTP | Provides margin during lossy encoding; measure the exported master |
| If mastering louder than −14 LUFS | Below −2 dBTP | Spotify advises extra peak margin to reduce encoding distortion |
| Premium playback options | −11 / −14 / −19 LUFS (Loud / Normal / Quiet) | Listener settings vary; do not optimize for only one mode |
| Delivered format | Native stereo FLAC preferred; WAV accepted | Send one highest-quality native stereo master, not multiple custom “Spotify versions” |
| Sampling / word length | 44.1 kHz or higher; 24-bit when native; 16-bit only when highest available | Retain native master resolution; no unnecessary up/downsampling before delivery |
References: Spotify loudness normalization and Spotify audio file formats.
Important distinction: −14 LUFS is Spotify's normalization/reference recommendation, not an upload acceptance threshold. A higher-loudness master is typically turned down during normal playback. True-peak safety, audible distortion and musical dynamics still matter.
Mastering protocol
- Receive an approved mix. Request a stereo premaster without unnecessary brickwall limiting, unless limiting is essential to the mix aesthetic.
- Listen for upstream problems. Fix vocal masking, clicks, stereo instability and damaged AI artifacts at the mix or source if possible.
- Set the artistic loudness. Level-match reference tracks before judging impact. Use LUFS-I, LUFS short-term and dynamics as context, not a race to a number.
- Control peaks carefully. Monitor true peak (dBTP) with reliable oversampling; audition limiting artifacts, transient loss and low-frequency pumping.
- Render once to the required native format. Keep the sample rate and bit depth of the native master when supported; dither once only for an actual fixed-point bit-depth reduction.
- Measure the rendered file. Record LUFS-I, max dBTP, sample peak, duration, channel count and any delivery exceptions.
- Audition an encoded preview. Check for overs, consonant grit, smeared cymbals and missing tails after a representative lossy transcode.
- Archive the native master and metadata. Preserve credits, permissions, source version and provenance.
A decision example
A punchy dance master measures −10 LUFS-I / −2.2 dBTP and passes an artifact check. On normal Spotify playback it will generally be attenuated toward the platform's reference; a forced reduction to −14 LUFS may or may not improve the musical result. Compare the two versions at the same perceived level before choosing. This is an illustrative engineering decision, not a recommendation to always master at −10 LUFS.
Common misconceptions
- “Every master must be exactly −14 LUFS.” No. The service publishes a reference, and playback gain is not the same as changing the master file.
- “A true-peak ceiling fixes bad mix balance.” No. It limits overs, not masking or collapsed transients.
- “EQ, saturation, resampling or limiting makes a track undetectable as AI.” There is no substantiated universal setting. These controls have ordinary sonic purposes; they should not be used to conceal origin or defeat provenance signals.
- “Lossless delivery means listeners always hear lossless.” Playback depends on eligibility, device and settings.
Delivery checklist
- One approved stereo master in native supported lossless resolution.
- Integrated LUFS and max dBTP recorded from the exported file.
- No unintentional clipping after encoding or conversion.
- No cut tails, wrong channel order, accidental silence or duplicate versions.
- Distributor's current requirements, credits, rights and AI disclosures confirmed.
- Source, master and measurement report archived.
See mix diagnostics, provenance and release, and detailed QC.