Recording Celtic Vocals at Home: Room, Microphone, Stacking and the Reverb That Does Not Drown It

The room is the microphone

Every conversation about home vocal recording starts in the wrong place, which is a shop. The microphone is roughly the fourth most important variable in how your vocal will sound, behind the performance, the room and the placement — and the room is the one people spend nothing on.

A small untreated bedroom has two problems. Early reflections from nearby walls arrive within a few milliseconds of the direct sound and comb-filter it, producing the hollow, boxy quality that no amount of EQ removes. And modal build-up in the low-mids makes small rooms boom around 100–250 Hz, exactly where a voice gets its body.

What actually helps, in order of return per pound spent:

• Sing into the room, not into a corner. Stand about a third of the way into the longest dimension, facing the long way. Free.
• Kill the first reflection behind you. A duvet on a clothes rail two feet behind the singer absorbs the loudest early reflection. Genuinely effective, genuinely free.
• Soften the wall in front of the singer. A bookcase filled unevenly with books diffuses beautifully. A heavy curtain works too.
• Carpet or a rug. The floor bounce is the reflection people forget.
• Then, if there is money left, two or four 100 mm mineral-wool panels at the first reflection points and straddling the corners.

Microphone choice for this voice

The ethereal style has a specific set of problems: a lot of quiet detail, a lot of breath, sibilance that becomes obvious at low levels, and an upper register that turns strident if the microphone has an aggressive presence peak.

• Large-diaphragm condenser. The default, for good reason: sensitive, detailed, flattering. In an untreated room, its sensitivity is also its problem — it hears the room as clearly as it hears you.
• Dynamic microphone. Genuinely underrated here. A good broadcast dynamic is far less sensitive to the room, handles proximity beautifully, and gives an intimate close sound. It needs more clean gain, so a quiet preamp or an inline booster matters.
• Ribbon. Naturally rolled-off on top, which flatters a bright or sibilant voice and suits this style. Figure-of-eight pattern means the null to the sides can be pointed at your worst reflection — a free acoustic treatment.
• Small-diaphragm condenser. Not usually for lead vocal, but excellent as a room pair for the choral beds described below.

The honest summary: a mid-priced condenser in a treated corner will beat an expensive one in an untreated room, every time, and the difference is not subtle.

Placement, and the rules that come from physics

Placement is where most home recordings are won or lost, and the variables are few enough to learn in an afternoon.

• Distance. For an intimate ethereal lead, 15–20 cm. Closer gives you proximity bass boost and mouth noise; further gives you more room and less intimacy. For a fuller, more classical sound, 30–40 cm — but only if the room deserves it.
• Height. Capsule at or slightly above nose height, angled down a few degrees. This aims the axis above the mouth and reduces plosives and sibilance before any processing.
• Off-axis. Rotate the microphone 10–20 degrees so you are not singing straight into it. Costs a small amount of top end, removes a large amount of harshness.
• Pop filter at 8–10 cm from the capsule, which also conveniently stops the singer drifting closer as the take goes on.
• Consistency matters more than perfection. Mark the floor. A singer who moves 15 cm between takes will produce stacks that never quite blend, and no amount of editing fixes it.

Levels and the thing nobody tells beginners

Record at 24-bit and aim for peaks around −12 to −6 dBFS. There is no advantage whatsoever to tracking hot at 24-bit — the headroom is free and clipping is permanent. Use a gentle compressor on the way in if you must (3:1, two or three dB of reduction at most), but if you are unsure, record flat and compress later. You cannot undo it.

Comping and tuning without sanding off the performance

Record three to five full passes before you edit anything. Sing the whole song each time; punching in phrase by phrase produces takes that are technically clean and emotionally discontinuous, and listeners hear the discontinuity even when they cannot name it.

Then comp — assemble the best version from the passes:

1. Listen to each pass all the way through and mark the phrases that are alive, not the ones that are accurate. Life first; accuracy is fixable and life is not.
2. Cut at breaths, not mid-word. The breath before a phrase belongs to that phrase and carries its attack.
3. Keep crossfades short, 5–10 ms, and place them in silence wherever you can.
4. Keep the breaths in. Removing them entirely is the fastest way to make an intimate vocal sound synthetic. Ride them down 4–6 dB if they are distracting.

On tuning: this style is especially vulnerable to over-correction, because a long, quiet, straight-toned note has nowhere to hide the artefacts. Use the slowest retune speed that fixes the problem, correct individual notes by hand rather than running the whole track through a fast setting, and leave the small drift at the start of a note — that scoop is a large part of what makes the line sound human. If the pitch is wrong enough that slow correction cannot fix it, sing it again.

Building a stack that sounds like one voice

The layered choral bed is the signature texture of contemporary Celtic and ethereal production, and there is a craft to it beyond pressing record five times.

Doubling the lead

Two passes of the same line, panned hard left and right, with the original lead in the centre. Do not copy and detune a single take; the entire effect depends on the natural micro-timing and micro-pitch differences between real performances.

The harmony bed

1. Record each harmony part twice and pan the pair left and right. Three parts recorded twice gives you six tracks and a genuinely wide bed.
2. Sing harmonies quieter than the lead — not just mixed quieter, actually performed more softly. A quiet performance has less high-mid energy and blends instead of competing.
3. Use a darker vowel. “Oo” and “oh” blend; “ah” and “ee” fight the lead for intelligibility. Wordless beds usually want the darkest vowel that keeps the pitch clear.
4. Vary the microphone distance by a few centimetres between layers. Slightly different room content per layer is what stops six identical tracks sounding like one phasey track.
5. Filter ruthlessly. High-pass every harmony track at 150–250 Hz. The low end of six vocal tracks is mud and it is doing nothing that the lead is not already doing.

Reverb as architecture

Reverb in this genre is not an effect applied at the end. It is the room the song is set in, and it should be chosen before the arrangement is finished.

• Use sends, never inserts. One or two reverb buses that many tracks feed, so everything sits in the same space. Separate reverbs per track produce a recording that sounds like it was made in four buildings.
• Two reverbs, different jobs. A short plate or chamber, 0.8–1.4 s, to give the voice a body and glue it to the arrangement; and a long hall, 3–6 s, used sparingly for scale. Send the lead mostly to the short one and the harmony bed mostly to the long one.
• Pre-delay is what keeps the lyric intelligible. 20–60 ms of pre-delay separates the direct sound from the tail, so the words stay in front of the wash. This single control does more for clarity than any EQ move.
• Filter the return. High-pass the reverb bus at 300–400 Hz and low-pass around 7–9 kHz. An unfiltered tail is where the mud and the sibilant hiss both live.
• Duck the tail under the words. A compressor on the reverb return, keyed from the lead vocal, pulls the wash down while the line is being sung and lets it bloom in the gaps. This is the trick behind almost every big, clear ethereal vocal you have ever admired.
• Tempo-sync a delay instead of adding more reverb. A dotted-eighth or quarter-note delay with the feedback low creates size without smearing, and it keeps the arrangement rhythmic rather than foggy.

Recording the instruments around the voice

An ethereal vocal needs something to sit in, and the instruments in this idiom each have a specific recording problem worth knowing before you press record.

Harp

The harp is a large radiating surface, not a point source, so a single close microphone will capture one region of the instrument and miss the rest. Two approaches that work in a home room:

• Spaced pair, one microphone at roughly 40 cm aimed at the upper strings and one at the lower, both about level with the middle of the soundboard. Pan them modestly — 40 per cent rather than hard — or the instrument will feel unnaturally wide behind a centred vocal.
• Single microphone at about 60 cm, on-axis with the middle of the soundboard. Less detailed, far more robust, and much easier to fit under a voice.

Watch for two noises the player stops hearing: pedal or lever movement, and fingernails on wire. Both are part of the instrument at a distance and become distracting up close.

Bodhrán and frame drums

One microphone in front of the skin at 20–30 cm captures the tone; a second behind, pointed at the player's inner hand, captures the pitch bends that make the instrument expressive. If you only have one microphone, take the front and accept that you lose half the vocabulary. High-pass at 45–60 Hz to remove the floor rumble that a frame drum transmits through a stand.

Strings

Violin and viola are bright and directional and will sound harsh close up in a small room. Back off to a metre if you can, aim slightly off the bridge, and let some room in. Cello is the opposite problem: it needs distance to develop, and the low end will excite your room's worst mode. If the cello sounds boomy, move the player, not the EQ — a metre in any direction usually fixes it.

Piano and harmonium

An upright piano recorded from behind, over the player's shoulders with the lid open, usually beats anything you will get from the front in a domestic room. A harmonium is largely pump noise up close; give it a metre and let the drone bloom.

A minimum viable chain, and the discipline that beats it

Here is the order I work in when mixing a lead vocal in this style. Every stage is optional; the order is not.

1. Clean-up. High-pass at 80–100 Hz. Manually ride out the worst breaths, clicks and chair creaks before any dynamics processing amplifies them.
2. Volume automation first, compression second. Draw the phrase-level balance by hand so the compressor is correcting texture rather than structure. Most home mixes do this backwards and end up with a vocal that pumps.
3. Subtractive EQ. Find the one boxy resonance around 200–400 Hz with a narrow band and cut two or three dB. One cut, not five.
4. Compression, gently and in two stages. 3:1 with a slow attack and a few dB of reduction, then a second, faster compressor doing one or two dB. Two gentle stages sound far more natural than one aggressive one.
5. De-essing after compression, because compression is what made the sibilance obvious in the first place. Target the specific band, usually 5–8 kHz, and use the least amount that works.
6. Additive EQ last, if at all. A wide, gentle shelf above 10 kHz for air. Resist the presence boost around 3 kHz; it is what makes home vocals sound shouty.
7. Sends to the two reverbs and the delay.

And the discipline that beats all of it: get it right at the source. The single best decision on the 2016 single “An Samhradh” was to record almost nothing — voice, piano and one floor tom — in a room that suited it, and let Mark Wyse mix something that already worked. There is no chain that rescues a performance recorded badly in a bad room, and there is no plug-in that adds conviction. Treat the room, place the microphone properly, sing it until it is true, and the mix becomes the easy part.

home recording vocal recording microphone technique reverb mixing studio production stacking