Updated 2026-10-11
The Stadium PA Filter does not make a bad recording good. It makes whatever you give it narrower, louder, and more echoing, and it does that to the room noise and the stumbles in your take exactly as it does to your voice. This guide covers the recording that happens before you open the tool: where to stand, how loud to speak, and what to leave out.
The advice below is general recording practice, tied wherever it matters to what the processing chain on this site actually does with your file. If you want the presets and sliders explained, the How It Works page covers them. Here the focus is the raw take.
The tool runs your audio through a fixed order of stages. Compression comes first, then a band-pass filter, an optional resonant boost, distortion, and finally the echoes and the tail. Each stage assumes the one before it handed over a clean voice. Anything that is not your voice, such as fan noise or room echo, travels through every stage alongside the words.
Compression is where background noise first becomes a problem. The compressor sets its threshold at 0.3 of full scale and only turns down the part of each sample above that line. It then applies makeup gain so the result is not quieter, and the higher the compression setting, the more makeup gain it adds. At the Ballpark preset's setting of 0.65, that gain works out to roughly 1.7 times. Quiet material below the threshold is not turned down, so it gets lifted with everything else, and a faint hiss in your room ends up louder next to your voice than it was when you recorded.
The echoes then multiply whatever is left. Ballpark sends a first reflection 90 ms after the dry signal at 40 percent of the dry level, before the wet mix is applied. Three more repeats follow, spaced 150 ms apart, each half the level of the one before. A cough, a lip smack, or a chair creak is repeated up to four times in the output. A clean take gives the echoes nothing to repeat except words.
Echo in the room where you record is the hardest problem to fix afterwards, because the tool adds a room of its own on top. A bare bedroom, a bathroom, or a kitchen gives short, hard reflections that the Ballpark and Arena presets then wrap in another reverb. Two rooms stacked together smear the consonants, and a PA read depends on crisp consonants more than most speech does. Record wherever you are surrounded by the most soft things.
A wardrobe full of clothes is a classic choice because hanging fabric absorbs reflections from every direction. A bed with a duvet, a sofa, curtains, and a rug do the same job at a smaller scale. In a bare room, hang a blanket behind the microphone and another behind your back, or speak into the open door of a clothes cupboard. Keep away from the corners of hard walls and from the middle of a large empty floor. Close the windows, because traffic is hard to separate from a voice once the two are on the same file.
Switch off whatever hums. Refrigerators, fans, air conditioners, and computers with loud cooling all put steady noise into the take. The Megaphone preset is the least forgiving of these, because it keeps only the band from 1,000 to 2,400 Hz and boosts the area around 1,500 Hz. That band is where speech carries most of its clarity, and it is also where hiss and buzz show up. The preset does not remove noise there. It is the part that survives.
Start with the microphone about a hand-span from your mouth and adjust from there. Closer than that, most directional microphones exaggerate the low frequencies, which is known as the proximity effect. The Ballpark preset cuts below 200 Hz, so some of that boom is removed, but the filter is a gentle first-order design that cuts at 6 dB per octave. It reduces low-end boom rather than deleting it, so it is better not to record the boom at all.
Aim the microphone slightly off-axis, at the side of your mouth or your chin, so the blast of air from p and b sounds does not land on the capsule. On a phone, hold it the same distance away with the microphone at the bottom edge pointed toward you, and keep your fingers off the opening. A laptop microphone is usually far from your mouth and picks up keyboard thumps and fan noise, so earbuds with a microphone or any headset will likely beat it. A USB desktop microphone sits in between. Put it on a stand or a stack of books so it does not carry vibration from the desk.
Keep the distance the same for the whole take. If you lean in for a loud line and drift back for a quiet one, the level swings, and the compressor reacts to those swings instead of to your delivery. A strip of tape on the desk, or a hand resting in the same place, is enough to keep you honest.
The guidance on this site is to aim for peaks around -6 dBFS, and the reason is concrete. After the compressor adds makeup gain, the signal passes through a tanh soft limiter and then the distortion stage. The drive factor of that stage is 1 plus five times the distortion amount. That is 1.5 at the Ballpark setting of 0.10 and about 5.75 at the Megaphone setting of 0.95. Every stage pushes a hot signal harder into its curve, so a recording that starts near full scale arrives at the end badly mangled.
Clipping in the recording itself is a separate problem, and it cannot be undone. When a microphone input is overloaded, the tops of the waveform are flattened, and the flattened samples no longer hold the information that was there. No later step can rebuild it, and the tool's distortion would add its own saturation on top of damage that is already baked in. If the meter in your recording app touches the top, even briefly, record the take again with the gain turned down or with your voice further from the microphone.
Too quiet is the lesser problem, though it is not free. At the end of processing, the tool normalises the output so that the loudest peak sits at 0.891 of full scale, which is about -1 dB. That step only ever boosts; it never turns a louder file down. A quiet take is therefore lifted in the last stage, and the noise rises with it. A healthy level at the start keeps the voice well clear of the noise, so there is less to lift.
Plosives are the bursts of air in words like Paul, big, and Boston. They hit the microphone as a low thump that can sound harmless in your headphones and ugly after processing. A pop filter helps, and a sheet of tights stretched over a bent coat hanger is a well-known homemade version. The Megaphone preset cuts hard below 1,000 Hz, with three filter stages giving about 18 dB per octave, so it hides most of a thump. Ballpark, with its gentle 200 Hz cut, reduces a thump without removing it.
Breaths and mouth noise are the other offenders. A PA read is loud and clipped, so a big inhale at the front of the clip sounds odd, and the echoes repeat it. Take your breath before you start the recording, or pause and breathe away from the microphone. Sip water first, because a dry mouth makes clicks and lip smacks. The narrow top ends of Vintage at 4,000 Hz and Classic at 3,500 Hz do not remove those clicks, since they reach well across the band.
Hard s sounds deserve a mention as well. They are bright and carry a lot of their energy near the top of the band. Modern, with its 6,000 Hz upper limit, passes most of that, while Classic trims it. If a take is harsh on s sounds, speak slightly across the microphone instead of straight into it, or record the line again. Choosing a preset to hide a sibilant take is a poor trade, because the preset also changes everything else about the sound.
A stadium announcer does not talk the way you do at dinner. The delivery is slower and more deliberate, with a clear gap between words, because the venue's own reflections blur anything spoken faster. The tool models that blur. Arena has a 2.2-second tail and Outdoor has a 2.5-second tail, so a quick read runs over itself and the last words arrive in the wash of the first.
Give every syllable a clear start. Hard consonants at the front of words, such as the t in 'tonight' or the k in 'Kowalski', give the compressor and the echoes a sharp edge to carry. Hold a beat between a name and a number, or between a name and a team, so that the repeats, which land 150 ms apart after the first one, fall into the gap and not on the next word. Let your pitch drop at the end of a line the way a real announcer's does.
The energy should come from your chest, not from raw volume. Speak firmly and with presence, but do not shout. Shouting pushes you toward clipping, and it makes the voice harsh in the narrow band that Megaphone keeps. Standing up while you record opens your chest and usually steadies the voice.
Leave a short moment of quiet before the first word and after the last. The tool already extends the output by the length of the reverb tail, 1.3 seconds for Ballpark, so you do not need to leave room for the tail yourself. What you do need is a clean start. A little silence lets you trim exactly, and a clip cut tight against the first syllable can sound chopped once the compressor has worked on it.
Record several takes in a row and do not judge them while you are still performing. Read the line three or four times with small changes in pace, then listen back. The best take is rarely the first one. Save each as a separate file so you can run the best two through the tool and compare them under the same preset.
The tool accepts WAV, MP3, M4A, and OGG files up to 50 MB. If your recording app offers a choice, record WAV, since it holds the audio uncompressed and MP3 does not. The effect keeps the sample rate and channel count of the file it is given, so a stereo file simply gets the same treatment twice and takes up twice the space. Mono is enough. On export, WAV is written as 16-bit PCM and MP3 at a constant 192 kbps.
Listen to the processed file on more than one playback device: a phone speaker, a laptop speaker, and headphones. Small speakers cannot reproduce much below 200 Hz anyway, so a Ballpark result sounds closer to a full-range recording on a phone than it does in headphones. Headphones reveal the clicks, breaths, and noise that the speakers hide. If the take only works on the small speaker, it will not hold up for everyone else.
The most common mistakes are effects stacked on top of effects. These include recording through a voice-changing app, switching on reverb in a streaming program, or leaving automatic noise suppression and gain control active. Those last two can make a voice pump and warble, and the tool then compresses the warble. Record dry, apply the effect once in the tool, and keep the dry file in case you want a different preset later.
Stacked reverb is the mistake a good take cannot recover from. Ballpark blends in 25 percent wet signal and Arena 42 percent, and those amounts were set with a dry recording in mind. Add your own room first and you push the total well past what the preset intended, while the consonants turn to mush. If a result sounds like a cave, the cause is usually in the recording rather than in the slider.