You treat it as two deliveries that have to end up in one loudness and one tonal frame, and you settle the technical questions before anyone argues about creative ones. Sample rate, bit depth, tempo map, tuning reference, vocal alignment and headroom on the mix bus all have to match across both camps, because a collaboration between two bands recorded in different rooms almost never arrives matched. I produced and mastered "In The Shadows Of Ukraine," released by Kalush Orchestra with the Finnish rock band The Rasmus on Columbia in October 2022, and the format itself is the interesting part.
What was the record, exactly?
"In The Shadows Of Ukraine" came out on 21 October 2022 on Columbia. The Qobuz release metadata from Sony lists my roles on it as producer and mastering engineer. NME covered the release at the time. On paper it is one song. In practice it is a Ukrainian rap and folk group with a sopilka player and a rap lead, working with a Finnish rock band whose sound is built on a rock vocal, distorted guitars and a drum kit that occupies the same midrange the sopilka lives in.
Nothing about that combination is impossible. It just has to be decided rather than mixed into existence at the end.
Why do two-band collaborations arrive as mismatched files?
Because each side records in the room it already trusts, with its own engineer, its own converters and its own habits. The mismatches are boring and consistent: one camp works at 44.1 kHz because that is what the label asked for last time, the other tracks at 48 kHz because the drummer's session template came from a picture job. One side prints vocals with compression and a de-esser baked in, the other sends raw. Tempo maps differ by a fraction of a BPM if anyone played to a click that was typed by hand rather than imported.
Set three things before the first file moves, in writing:
1. Sample rate and bit depth for everything. 48 kHz, 24-bit is the safest single answer, because it is already the video standard and downsampling to 44.1 kHz later is one controlled step rather than a surprise. 2. One tempo map and one bar-one reference. All stems start at the same timecode, including the silent ones. A stem that starts at the first note instead of bar one costs an hour of nudging on the day you have the least time. 3. Tuning reference A=440 Hz, stated out loud, because folk instruments do not all agree with a pitch reference by default and a sopilka against a slightly sharp guitar reads as a bad take rather than a tuning issue.
That list looks trivial written down. Skipping it is what turns a two-week collaboration into a five-week one.
How do two lead vocalists share one arrangement?
By giving them different registers, different reverbs and different halves of the arrangement, not by fighting for the same 2 kHz.
A rap lead and a rock lead ask for opposite things. Rap wants consonants intact, tight low mids and dry placement so the rhythm reads. Rock wants sustain, a tail, and enough low mid weight to sound like a chest. Put both through the same chain and the rap goes small while the rock vocal goes muddy.
What works: separate reverb sends, not a shared plate. Short room of 0.6 to 0.9 seconds on the rap vocal so consonants stay in front, a longer hall on the rock vocal that lives at least 20 ms later so the two never smear into one tail. High pass the rap vocal higher than instinct says, often 110 to 130 Hz, because the weight it needs comes from the arrangement rather than from its own low end. Then carve the guitars, not the vocal: a 2 to 3 dB dip in the guitar bus around 2.5 to 3 kHz opens the exact band where diction lives, and it is inaudible on the guitars in context.
The folk instrument needs the same treatment in reverse. A sopilka sits high and cuts through anything, which is why it usually needs dynamic control rather than EQ carving. Duck it 2 to 3 dB against the lead vocal instead of turning it down statically, and it stays present in the gaps without competing during lines.
What does the master have to protect in a record like this?
The transition points. The verse-to-chorus move between a rap section and a full rock chorus is a much bigger jump in spectral density than a normal pop song ever asks for, and that is where a heavy-handed master audibly pumps.
Practical settings I hold to on this kind of material. Bus compression at the master stage stays under 1.5 dB of gain reduction, with a release long enough to ignore the snare, usually 300 ms or auto. Limiting is decided by the loudest chorus, not the average, and if the chorus needs more than 3 dB of limiting to reach the target, the mix comes back rather than the limiter working harder. True peak ceiling at -1 dBTP so lossy encoding does not clip on playback, which is the same ceiling Apple Digital Masters asks for in its delivery specification. Integrated loudness is measured to ITU-R BS.1770 and set for the destination, not for the meter, because every major streaming service normalizes playback anyway.
One more thing specific to bilingual records: check both lyrics at low playback level, around 55 to 60 dB SPL, and check them with a listener who does not speak the language. If the shape of the line survives at conversation volume to someone who cannot parse the words, the vocal balance is right. A mix that only works loud is a mix that only works for the people who made it.
What do you actually deliver at the end?
More than the master. A record with two bands, two publishers and two territories generates paperwork immediately, so the file set should answer it before the questions arrive: the master at 24-bit, an instrumental, a version without the lead vocals for television, and stems grouped consistently from a common start point. Everything named by content, everything at the same sample rate.
The split sheet matters more here than on a single-artist record, because two bands means at least two publishers and often two performing rights organizations in different countries. Write it the week of the session, not the week of the release.
What to do with this
Sending a collaboration to master? Send both camps' stems from a common start point at 48 kHz, 24-bit, with the tempo map, and you get a delivery date and a list of anything that will not line up on the same day. The verified credit list is on the [credits page](/credits.html), and more on how I work is on the [about page](/about.html).