Send or insert. It’s the oldest argument in mixing, and it still trips people up because both answers are correct depending on what you’re trying to do.
I’ve mixed enough techno and house to say this plainly: the reverb you put on an insert is a sound design tool. The reverb you put on a send is a mixing tool. Confuse the two and you get either a wash that eats your low end or a tail that refuses to sit anywhere in the stereo field.
Let’s separate them properly.
What an insert reverb actually does to your signal
When you drop a reverb directly on a channel, the plugin becomes part of that sound. The dry signal and the wet signal are locked together through one fader. You can’t rebalance them without opening the plugin, and anything downstream of that channel, EQ, compression, saturation, now processes the reverb tail along with the source.
That’s the whole point. On a stab, a dub chord, or a single hit you want to smear into the track, an insert reverb commits the effect to the sound. Turn the mix knob to 40% and you’ve made a new instrument. This is where character reverbs shine.
Valhalla VintageVerb is my usual pick here. The 18 modes cover the classic Lexicon 224, EMT 140 plate, and AMS RMX16, each with the decay artifacts and frequency response of the original hardware, and the Color control lets you dial in how much vintage grit you want. It’s also light enough on CPU to run several instances without your session complaining. That matters when you’re using reverb as tone rather than as space.
Convolution reverbs behave differently. They fire an impulse response, a measured profile of a real room or hardware unit, at your signal, so they reproduce actual spaces with realistic results. On an insert, that can be useful for one specific job: making a sampled break sound like it was recorded somewhere. Load an IR of a small room and suddenly your drum loop has a location.
The catch. Convolution tails are static. You can’t easily shape their decay the way you can with an algorithm, and they don’t respond to modulation. For a static ambient bed, fine. For anything that needs to move, you’ll fight it.
Why sends are the default for electronic mixes
A send splits your signal. The dry channel stays untouched, and a copy goes to an aux bus with the reverb on it. Now you have one reverb feeding many sources, and each source has its own send amount.
Three reasons this wins in most electronic sessions.
- Shared space. Kick, clap, hats, and a vocal all hit the same reverb. They sound like they occupy the same room instead of four separate ones.
- Control without commitment. You automate send levels, so the snare tail can be long in the breakdown and short in the drop without touching the plugin.
- CPU and clarity. One instance instead of eight. One place to EQ the tail instead of eight.
That last point is where most people mess up. The tail is part of the mix now, so it needs processing. If your send reverb has no filtering, you’re dumping low-frequency energy into a busy mix and wondering why the low end turns to mud.
The frequency control question
This is the real reason I reach for one reverb over another on a send.
FabFilter Pro-R was built around this problem. The interface is a spectral display showing exactly where the reverb energy lives, and you can EQ the reverb from that display using the same engine as Pro-Q 4. The 6-band EQ lets you carve mud below 200 Hz, tame harshness around 3 kHz, and add air above 10 kHz without leaving the plugin. For a send bus in a dense techno mix, that’s not a luxury. Darken the reverb below 500 Hz and a vocal or lead stays present while the tail still fills the stereo field.
Pro-R runs about $199, often discounted to around $129 in FabFilter sales. It’s clean and surgical rather than characterful. If you want the sound of a specific vintage plate or hardware unit, this isn’t it.
Valhalla Room takes the opposite approach. It’s around $50, has excellent room modes on snare and drum buses, and hall modes that push instruments back in a mix. But it lacks surgical frequency control, so precise EQ integration means a separate EQ plugin after it. Same value argument as VintageVerb. Different job.
For pure character on a send, Relab’s LX480 emulation is the one I keep coming back to on vocals. It has a way of putting a track in a smooth space without conflicting with other elements. The trick with a long pre-delay to push the tail out of the way of the dry vocal is standard practice for a reason. It works.
Where sends go wrong
Two mistakes kill more electronic mixes than anything else.
First, no pre-delay. Reverb that starts immediately blurs the transient. Push the pre-delay to 20 to 40 ms on percussive sources and the attack stays sharp while the tail still fills in behind it.
Second, too many sends. If you’ve got a plate, a hall, a room, and a spring all running, you’ve got four different spaces fighting. Pick two. One short for glue, one long for depth. That’s usually enough for a full track.
And watch the stereo width. A send reverb that’s fully wide will smear your mono compatibility. Check in mono before you finish.
When to break the rule
Insert reverb is the right call in three cases, and I use all of them.
One, sound design. A metallic stab drenched in an algorithmic reverb with a long decay isn’t a mixing decision, it’s the sound. Commit it.
Two, a single element that needs its own space no other sound shares. A lone vocal or lead that should feel intimate while everything else is dry.
Three, when you’re resampling. Printing a reverb into an audio file locks the character in and frees the CPU. That’s a workflow choice, not a mix choice.
Everything else? Send it. Set up two aux buses, filter the lows on both, use pre-delay, automate the levels, and stop thinking about reverb as an effect. It’s a bus. It’s part of your mix architecture.
The plugins matter less than people think. You can build a finished techno track with one decent algorithmic reverb on a send and a free option like Valhalla Supermassive for the long tails. What matters is knowing whether you’re adding space or making a sound. Get that straight and the send versus insert question answers itself.