Stem mastering and delivery formats (WAV, dithering, ISRC, DDP, Dolby Atmos)

Stem mastering

Instead of a single stereo file, mastering works on 4–8 stereo stems (mix groups). It allows corrections that would be impossible on the full mix (lowering the vocal by 0.5 dB, controlling the 808 without touching the kick, de-essing only the vocals), at the cost of more time and part of the mix's responsibility shifting to mastering.

Typical stems

Export rules

• All stems from the same start point (bar 1 or 0:00, even if the stem starts later) and with the same length (up to the end of the last tail).
• Same sample rate and bit depth as the mix (24 bit or 32 bit float).
• Without the master chain (no limiter, glue, master EQ); bus processors can stay if they are part of the sound (agree on this).
• Reverbs and delays: either inside each stem (each with its own returns: complex), or on a dedicated FX RETURNS stem.
• Null test: the sum of the stems, with a flat master, must match the premix (when you invert the phase of the sum against the premix, the residual must be below about −60 dBFS). If the mix bus glue was active in the premix, the sum won't null completely: flag it.
• In FL: export with the Split mixer tracks option (renders each mixer track to a separate file); organize the mixer so the stems coincide with the buses; turn off the master effects (the master effects option in the render window) or bypass them; Tail: Leave remainder so the tails aren't cut off.
• Clear file names: "Title_STEM_DRUMS_24b_48k.wav".

Audio formats

Dithering

• What it is: very low-level noise added when you reduce the resolution (e.g. from 24 or 32 bit to 16 bit), to replace quantization distortion (audible in reverb tails, fades and quiet passages) with an almost imperceptible hiss.
• When: only when reducing bit depth; only once; as the last process. After dither you no longer change gain, sample rate or anything else.
• Types: TPDF (flat, sufficient at 24 bit), with noise shaping (moves the noise toward less audible frequencies; useful at 16 bit). It isn't needed when exporting to 32-bit float.
• 16/44.1 kHz workflow (e.g. CD): high-resolution master → high-quality sample rate conversion (SRC) to 44.1 kHz keeping 32 bit → 16-bit dither (noise shaping optional) → export, with no further changes.
• In FL: use the export dither (if you export at 16 bit) or a dither in the last master slot (Ozone, Pro-L 2, MAutoDither) with the export dither turned off: never both.

Sample rate conversion

Convert with the highest-quality algorithm (in FL the highest resampling quality in the export window, e.g. 512-point sinc). Don't convert back and forth; don't upsample a master that was made at 44.1 kHz "to get a hi-res file".

ISRC, UPC/EAN and metadata

• ISRC (International Standard Recording Code): identifies a recording, not the composition (that has the ISWC). 12 characters, written as CC-XXX-YY-NNNNN: country (2), registrant (3), year of assignment (2), recording number (5). Example of the structure: IT-XXX-26-00001.
• Every different version (radio edit, remix, instrumental, extended version, remaster) has its own ISRC.
• ISRCs are assigned by national agencies (under IFPI supervision) or by an authorized registrant; in independent practice they are often supplied by the digital distributor (DistroKid, TuneCore, CD Baby, Believe…) at upload.
• UPC/EAN: the code of the release (album, EP, single), not of the track.
• Metadata: title, artist, featuring, authors, composers, producers, publishers, label, year, genre, explicit lyrics yes/no, language, credits. In the DDP and CD-Text, titles and artists for each track are included too.
• Keep a delivery sheet with: title, ISRC, duration, integrated loudness, true peak, sample rate/bit depth, versions, notes.

DDP (Disc Description Protocol)

• Standard format for delivering a CD premaster to the plant: a folder with the audio image (.DAT file), the identifier (DDPID), the stream descriptor (DDPMS) and the subcode descriptor (PQDESCR), often with a text file of titles and times and a checksum.
• It contains track order, gaps (pregap), index points, ISRC per track, CD-Text, EAN/UPC code.
• It is created with authoring software (Steinberg WaveLab, HOFA CD-Burn.DDP.Master, Sonoris DDP Creator, DSP Quattro). FL Studio does not create DDPs.
• Verify the DDP with a DDP player (listen to the entire disc, check the times and the checksum) before sending it.

Versions to deliver

• Main (primary master), Clean/Radio edit (without explicit words), Instrumental, A cappella (vocal only with its effects), TV track (instrumental + backing vocals), Stems for sync, remixes or live, extended/club version if needed.
• Same master chain for all the versions (with small loudness adjustments if the absence of the vocal requires it), same length and alignment so they can be swapped easily in video.

Dolby Atmos: basics

• What it is: an immersive format based on beds (fixed channels, e.g. 7.1.2) and objects (sounds with a position in 3D space, position metadata) up to 128 elements in total. The Renderer adapts the mix to any system: speakers (5.1, 7.1.4, soundbars) or binaural for headphones (Spatial Audio on Apple Music, Tidal, Amazon).
• Workflow: start from the stems/tracks of the stereo mix and place them in a 3D space in a DAW with Atmos support (Logic Pro, Pro Tools, Nuendo, Cubase, DaVinci Resolve, Dolby Atmos Renderer). FL Studio doesn't currently offer an integrated Atmos workflow: export stems and tracks and work in a compatible DAW (check the updates for your version).
• Delivery (Apple Music, commonly cited guidelines): ADM BWF file compliant with the Dolby Atmos Master profile, 48 kHz / 24 bit, integrated loudness no higher than −18 LKFS, true peak no higher than −1 dBTP, binaural render modes (Off/Near/Mid/Far) set for beds and objects, timecode at 24 fps, and the matching stereo version delivered separately (Atmos doesn't replace the stereo master). Updated requirements: official guides from Apple and Dolby and from the distributor.
• Atmos mix: lead vocal usually in the center (center channel or center object), bass in the bed, width and height with pads, reverbs, backing vocals, effects; watch the binaural translation (headphone listening) and the stereo downmix.

Common mistakes

• Stems that are misaligned or of different lengths.
• Dither applied twice or before gain changes.
• The same ISRC for different versions.
• 16-bit masters sent to platforms that accept 24 bit (you lose quality needlessly).
• MP3 sent as a master.
• Tails cut off at export (Tail not set).

How to verify

• Null test of stems vs premix.
• Full relisten of every final file (start, end, silences, clicks).
• Measure every file (LUFS, true peak) and compare with the delivery sheet.