How a voice changer reaches a Discord call
Discord does one simple thing with audio: it listens to whatever device you pick as its input, and it sends that sound to the call. It never asks what happened to the sound before it arrived. That is the whole trick. A voice changer that works in calls transforms your microphone in real time and then presents the result to Windows as a new input device. You select that device in Discord instead of your physical mic, and from that moment everyone in the call hears the character, not you.
This guide builds that chain with Patchd, a Windows virtual audio mixer, and Persona, its built-in voice changer. The same five steps apply to any voice app with a device picker, so once you have this working for one call app it works for all of them. If the idea of a voice changer is new to you, what is a voice changer covers the fundamentals first.
Step 1: pick a voice in Persona
Persona offers two kinds of voices, and the difference matters more in a call than anywhere else. Instant voices are built from fast DSP effects, real-time processing that reshapes your pitch, formant, and tone as you speak. They add no perceptible delay, so the rhythm of a conversation survives intact. AI voices rebuild your speech into an entirely new voice from the ground up. They sound dramatically more convincing, and they cost about a quarter of a second (roughly 245 ms) of delay, plus a few seconds of warm-up the first time you switch to one. After that first warm-up, switching between voices is near instant, and the new voice fades in rather than clicking in.
For the back-and-forth of a casual call, start with an instant voice. Save the AI voices for bits, storytelling, and moments where realism beats snappiness; more on why in the latency section below. Persona ships 111 voices, all original characters, and the free tier includes 6 rotating taster voices so you can find one that fits before paying anything. The full comparison lives in DSP vs AI voice changing.
Step 2: clean your mic before you change it
A voice changer transforms everything it hears, and that includes your fan hum, keyboard clatter, and room echo. Feed Persona a noisy signal and the noise gets rebuilt into the character right along with your words, which is exactly as strange as it sounds. So clean first, transform second.
In Patchd, your mic is a strip in the mixer with its own effects rack, filled from a + Add Node menu. Add AI noise suppression to turn steady background noise far down, then a noise gate to keep the mic silent in your pauses, and let Persona run as the final stage after them. Both cleanup nodes are free on every input channel. If you want the settings walked through, how to set up a noise gate and how to reduce background noise on your mic cover it in detail.
Step 3: route your mic strip to a bus
Your voice is now clean and transformed inside Patchd, but Discord cannot see inside a mixer. It needs a device. That device is a bus: a virtual output you assemble for one job, which other apps see as a microphone. On your mic strip, click the Bus 1 pill under the BUS header. It lights up cyan, and that lit pill is the entire routing step: your processed, character-voiced mic now flows onto Bus 1.
Step 4: select the bus as your input device in Discord
Now the step people miss. Open your voice app's audio settings, in Discord that is the voice and video section, and find the input device dropdown. It currently says the name of your physical microphone. Change it to Patchd Bus 1. That single dropdown is the handoff: Discord stops listening to your raw mic and starts listening to the finished chain, cleanup and character voice included, and everyone in the call hears the character speak from your tile.
If you skip this step, everything upstream is wasted. Your voice changes perfectly inside Patchd while Discord keeps pulling from the untouched physical mic, and the call hears plain, unprocessed you. Whenever a voice changer seems to work in the app but not in the call, the input device picker is the first place to look.
Step 5: put every voice on a hotkey, and keep one to bail out
Mid-call is the worst time to alt-tab into a mixer. In Persona, switching is hotkey-first: you bind a global key to each voice you use, and, crucially, one to Toggle Persona on/off. That toggle is the one that saves you: when the joke lands and the conversation turns serious, or someone calls who was not in on the bit, one keypress drops the voice and hands you back your own, without ever leaving your game. A new voice fades in from silence rather than switching on at full level, so there is no pop or click when you change characters, and the keys work from any window.
Step 6: hear yourself
The last step is monitoring, and it is worth doing. A character voice changes your timing and your delivery, and hearing the output is how you keep it sounding natural instead of rushed. In Patchd, route your mic strip to your headphone hardware output as well as Bus 1, so the same processed voice the call hears also reaches your ears. Patchd runs over ASIO and WASAPI Exclusive and adds about 10.7 ms through the bus at a 512 buffer, and smaller ASIO buffers take it lower, so with an instant voice your own monitoring feels immediate. If you hear yourself doubled or echoing instead, fixing mic echo sorts out the usual routing cause.
The honest latency picture for calls
Streaming forgives delay; a phone-style call does not. Instant DSP voices add effectively nothing, so a conversation with one flows exactly like a conversation without one. AI voices add roughly 245 ms, and in a fast back-and-forth that gap is real: you will occasionally start talking over someone because your last sentence arrived a beat late. For a bit, a story, or a one-liner delivered to the whole call, nobody cares and the payoff is worth it. For an hour of tight conversation while you play, an instant voice is the better default, with an AI voice one hotkey away for the moments that earn it. The deeper mechanics are in voice changer latency, explained.
Where Patchd is today
Patchd has not launched publicly yet, but you can install this chain today: it is in open beta, with every feature unlocked while the beta runs. The free tier described here, your mic strip, the cleanup nodes, sixbuses, and Persona's 6 taster voices, is what it will include at launch. If you want a feel for a character voice on your own words right now, the online voice changer runs in your browser today, and the browser mic test will confirm your raw mic is healthy before you build anything on top of it. You can try Patchd now in the open beta.