The short answer
A delay effect takes an incoming sound, holds a copy of it for a set amount of time, and then plays that copy back. Feed the copy partway into itself and you get a string of repeats, each one a little quieter than the last, until they fade away. Set long, that is an echo, and a delay is the tool that produces it on purpose. Set short, the same tool stops sounding like a repeat at all, which is the part most people never hear about. You hear it on a vocal that bounces the last word of a line, on a guitar that rings out in rhythm, and on the classic call-and-answer trail behind a lead.
The key word is on purpose. A delay effect is something you add because you want the repeats. That makes it the opposite of latency, which is delay you never asked for: the small lag between a sound going into your system and coming back out. One is a copy you add on purpose, the other is a lag on everything that you try to keep as small as possible. This page explains the effect, then draws the line between the two clearly.
How a delay effect works
Every delay is built from the same three controls, and once you can name them you can dial in any echo you have ever heard.
- Delay time. How long the effect waits before it plays the copy back, usually set in milliseconds or locked to a musical note. A time of 60 to 120 ms gives a tight slapback, a quick copy tucked right behind each word. A longer time of 300 to 500 ms gives a distinct, separate repeat you hear as its own event. Go shorter than a slapback and the copy stops sounding like a repeat at all, which the next section explains.
- Feedback. How much of the repeat gets sent back into the delay to be repeated again. Zero feedback gives a single echo. Raise it and you get a train of repeats that decays over several steps. Push it near the top and the echoes can ring for a very long time.
- Mix. The balance between your untouched sound, called the dry signal, and the echoes, called the wet signal. Low mix tucks the repeats politely behind you. High mix pushes them forward until the echoes are as loud as the source.
The clearest way to picture a delay is to send it a single short sound and watch what comes out. You get the original hit at the start, then evenly spaced copies marching off to the right, each shorter than the one before as the feedback bleeds away. The spacing is the delay time. The rate at which the copies shrink is the feedback.
Because the repeats are copies of the original placed later in time, a delay is one of the simpler effects to reason about. It does not reshape the tone the way an EQ does or squeeze the dynamics the way a compressor does. It just hands you back what you gave it, spaced out and fading, which is why it reads so instantly as an echo.
One knob, four sounds
Set feedback to zero, so there is exactly one copy, and mix to half, so that copy is as loud as you. Now delay time is the only thing moving, and it does not give you one effect at four lengths. It gives you four different sounds, because what changes is what your ear does with a copy that arrives that late.
At a few milliseconds the copy and the original cancel each other at evenly spaced frequencies, the comb a flanger is built from. At 3 ms those notches sit about 333 Hz apart, wide enough for your ear to hear as a change of tone color, so the voice turns hollow. Around 20 to 30 ms the notches are packed too tightly to resolve, and a copy that late is what a wall a few meters away would send back. Your hearing fuses reflections that close into the sound that made them, so you hear one thicker voice. Near 100 ms the copy is past that fusion window, which for speech closes somewhere around 40 to 50 ms, so it becomes its own sound: a slapback, landing inside the word you are still saying. Only past a couple of hundred milliseconds, roughly one syllable of speech, does the copy land after your next syllable and read as a repeat. That last one is the echo. Decide which of the four you want before you reach for the knob.
Delay is not latency
This is the confusion the title promises to clear up, because both words describe a sound arriving later than it started. The difference is intent, and it matters.
- A delay effect is wanted. You choose it, you set its time, and you want to hear the repeats. Turning it off makes the sound worse, not better, because the echoes were the point.
- Latency is unwanted. It is the lag your gear and software add just by processing the audio, and nobody asks for it. Turning it down always makes things feel better, because the sound arrives closer to when it happened.
The scale usually differs too, but scale is not the real line. An echo lives in the hundreds of milliseconds, and good monitoring latency sits in the single digits to low tens, small enough that you never notice it. Yet the thickening delay above runs at 25 ms, the same length as a sluggish monitoring path. What separates them is what stays on time. A delay effect adds a late copy while your original arrives exactly when it should; latency moves the whole sound late and leaves nothing on time. If you could hear your own voice echo back a quarter of a second after you spoke, that would not be a nice effect, it would be a broken, distracting lag. For the full picture of the unwanted kind, see what is monitoring latency and how a smaller buffer size lowers it.
Delay versus reverb
Delay and reverb are cousins, and they are easy to mix up because both add a sense of space and both are made of a sound repeating. The difference is how many repeats there are and how far apart they sit.
A delay gives you a countable number of distinct echoes with clear gaps between them, like a hand clap bouncing off a distant wall. Reverb gives you thousands of tiny reflections packed so tightly and so quickly that they blur into a single smooth wash, like the tail of sound in a large hall. Turn a delay time very short and pile the feedback high and it starts to smear toward reverb, which is a hint that the two really are the same idea at different densities. In practice you reach for delay when you want rhythmic, audible repeats, and reverb when you want a room around the sound.
Delay in Patchd
Patchd is a Windows virtual audio mixer, and its Studio voice-fx rack includes a delay effect you can add to a voice. You set the delay time and the feedback, and Patchd repeats your voice back as spaced, fading echoes, the deliberate creative kind described above. It is one node you can drop into a chain alongside a noise gate, an EQ, and a compressor, and where it sits in that chain changes the result, so it is worth reading effect chain order before you stack effects.
Here is the honest distinction that matters. The delay effect is something you add on purpose, and it can push a sound hundreds of milliseconds into the past because that is what makes the echo. That is completely separate from the latency of the engine itself, the unwanted lag as audio passes through Patchd. The engine is real time and runs over ASIO and WASAPI Exclusive, the low-overhead Windows audio paths, so at a small ASIO buffer its added latency stays in the single-digit milliseconds, engine internal. Adding a delay effect never changes that floor. You are choosing to hear repeats, not paying a hidden tax on the whole signal.
The delay lives in the Studio rack, but the underlying engine is identical on both tiers, so the low-overhead audio path is the same whether or not you run the voice effects. The per-channel DSP rack of a gate, EQ, and compressor is on every channel for free.
The takeaway
A delay effect is a deliberate tool that repeats your sound back as spaced, fading echoes. Delay time sets how far apart the repeats sit, feedback sets how quickly they fade, and mix sets how loud they are against the original. None of that is latency: an effect adds a copy you chose to hear, while latency is unwanted lag on the whole signal that you keep as small as you can. And the time does more than space the repeats: short settings make a voice hollow or thicker rather than echoing at all. Learn to hear the three controls and you can shape any echo, from a tight slapback to a long, ringing trail.
Patchd is in open beta, with a free per-channel effects rack and a Studio voice-fx rack that includes delay, both driven by the same real-time engine. If you want a Windows mixer that keeps its own latency low while giving you deliberate effects when you want them, join the waitlist and get notified when it ships. The Studio tier is $5.99 a month, $39.99 a year, or $119.99 once for a lifetime license, and the engine is identical on both tiers.