A single is one decision. An album is eleven decisions that only exist in relation to each other. On a single you can chase the loudest, widest, brightest version of that song and nothing argues with it. On an album, track 6 is heard three seconds after track 5, so its brightness gets judged against track 5 rather than against a reference, and the running order changes what each master should be. The delivery is different too: a single needs one WAV, an album usually needs a DDP image with PQ points, ISRCs and a barcode inside it.
What is the first thing that changes, and it is not the EQ?
The order.
Sequencing is a mastering decision that gets treated as an artist decision, and it is both. Two tracks in the same key, at similar tempo, with the same instrumentation in the top octave, sound like one long track no matter what I do to them individually. Two tracks with a 40 BPM gap and different vocal registers sound like a cut even with five seconds between them.
I ask for the order before I open a single file, then master to that order. If the order changes later, the level and tone decisions around the change point get revisited, because they were made for a neighbour that is no longer there.
How do you match levels across an album without flattening it?
Not by matching the integrated loudness numbers.
If every track lands at exactly the same LUFS integrated, a ballad that is meant to feel small will feel exactly as big as the single that opens the record, and the album loses its shape. Loudness measurement gates by level, which is why a sparse arrangement can measure the same as a dense one and sound obviously quieter.
What I do instead: master each track on its own merits, then play the whole record start to finish at one monitor level and only write down where the jumps hurt. The corrections out of that pass are usually 0.3 to 0.8 dB, three or four of them on a ten track album. Anything larger than about 1.5 dB says the master is wrong rather than the level.
The one number worth keeping in view is the spread. A record where the quietest track sits 2 to 3 dB under the loudest reads as dynamic. A record where the spread is 6 dB reads as unfinished, because listeners reach for the volume control. I measure with ITU-R BS.1770, the algorithm underneath every meter that says LUFS or LKFS, then ignore it during the listening pass.
What do the gaps actually do?
They control tempo at a scale above the songs.
A two second pause is the Red Book default, and a default is not a rule. On a record that runs hard, 1.2 seconds between two uptempo tracks keeps the pressure on. After a long fade, four seconds stops the next intro feeling like an interruption. Between the last two tracks I will sometimes go to six or seven, because that gap does the same job as a paragraph break.
Gaps also interact with fades. A fade that reaches true silence 800 ms before the track marker means the listener hears 2.8 seconds of nothing, not 2. I trim the tail and set the gap separately, so the number in the PQ sheet matches what the ear gets.
Does every track get the same chain?
No, and starting from a template is how albums end up sounding like one long track with different melodies.
What stays constant is the destination, not the route. If the record has a tonal signature, usually something in the low mids and something in the air band, that signature is applied per track by ear at whatever amount that track needs. On a mix that already has a lot of 8 to 12 kHz from a bright vocal chain, matching the album signature might mean 0.5 dB where the previous track took 2 dB.
The one thing I keep identical across an album is the final limiter and the dither. Changing limiters between tracks changes the character of the transient response, and that is audible across a cut in a way an EQ difference is not. Dither gets applied once, at the end, at the final word length. Dithering twice is the most common damage in second-opinion masters, and it happens because someone rendered a 16 bit file, then reopened it and rendered again.
What is a DDP image and why does the plant want one?
It is the whole disc as data, rather than a folder of audio files with instructions in an email.
A DDP image, in practice version 2.00, holds the audio as one continuous stream plus map and subcode files that say where every track starts, how long each pause runs, and what identifiers travel with each track. The format documentation lists the parts: an identifier file, a map file, the subcode descriptor and the audio stream. The plant reads it and presses, and nobody retypes anything, which is the point, because retyping is where track titles acquire spelling mistakes.
Red Book audio is 16 bit, 44.1 kHz stereo, with a maximum programme time just under 80 minutes, so the 24 bit masters are not what goes in the image. They stay in the archive folder, and the conversion happens once, with dither, on the way into the DDP. Even for a record that will never be pressed, building the image is worth an hour: it forces every piece of metadata into one place and gives you a checksum, so the version that arrives is provably the version that left.
Where do ISRCs come from and who assigns them?
The label or the artist gets them, not the mastering engineer, and they need to exist before the image is built.
An ISRC is 12 characters in three parts: a five character prefix, two digits for the year the code was assigned, and a five digit designation code chosen by the registrant. IFPI publishes the structure, and the important detail for a session is that the year element is the year the code was assigned to the recording, not the year of release. People fill it in wrong constantly, then wonder why two recordings share a code.
One code per recording, and a remaster or a different edit of the same recording gets its own. If the album version and the single edit share an ISRC, the streaming numbers for both go into one bucket and cannot be separated afterwards.
What does an album delivery look like when there is no CD?
Three sets of files, and a text document that ties them together.
The DDP image for replication if there will be one. A folder of individual 24 bit, 44.1 kHz WAVs for digital distribution, named with track numbers so they cannot be sorted wrong. And the archive folder with the unlimited masters at full resolution, which is what a vinyl cut needs later, because a vinyl master starts from a version that has not been limited flat.
The text document carries the running order with timings, every ISRC, the UPC, and one line per track for anything the plant or distributor needs to know. Fifteen minutes, and it removes most of the phone calls.
Frequently asked
Should I master the album and then the single from the same session? Master the album first, then revisit the single. A single competing on its own usually wants half a decibel more and a slightly firmer low end, and that version does not belong on the record.
How much does the running order really change? Enough that I ask for it twice, once at the start and once before the image is built. On about half the records I have worked on, at least one track moves.
Can I supply mixes at different sample rates? You can, and it costs you. Keep the whole record at one rate, ideally the rate the sessions were tracked at, and let the conversion happen once at the end rather than eleven times in the middle.
If you have a record and a running order
Send the mixes at full resolution, the order, and one sentence per track about what it is doing on the album. I will come back with where the level jumps are, which gaps need to move, and whether the order is fighting you. Full credit list at /credits, background and contact at /about.