# Make a custom voice effect from a description, like a walkie-talkie in a canyon

Turn a description like "walkie-talkie in a canyon" into band, grit, room and noise settings, put them in the order that places the listener, and keep the result as a reusable style.

Author: Kleo Ku

Canonical: https://voyager.so/blog/custom-voice-effects

Published: 2026-11-07T15:55:00-08:00
Updated: 2026-11-07T15:55:00-08:00

No preset is called "walkie-talkie in a canyon" or "through a wall at a party". You build those sounds from three parts: what the voice comes through, where the listener is, and how worn the recording is. Here are three, each built from a clean take in that order.

![Three illustrated scenes, each with a slider panel above it: a walkie-talkie on a canyon ledge, a door to a party in the next room, and an answering machine on a table. In each, the line plays clean, then the Voice sliders move, then the Room and Texture sliders, while a waveform and the spoken words fill in below.](https://cdn.moda.app/voyager/blog/custom-voice-effects/medley-v2-poster-4f17e92a9c71ce75.webp)

[Video: Three illustrated scenes, each with a slider panel above it: a walkie-talkie on a canyon ledge, a door to a party in the next room, and an answering machine on a table. In each, the line plays clean, then the Voice sliders move, then the Room and Texture sliders, while a waveform and the spoken words fill in below.](https://cdn.moda.app/voyager/blog/custom-voice-effects/custom-styles-medley_v2-4929a4e6f28d8d85.mp4)

Three custom styles, each built in two steps: the Voice sliders make the device, then Room and Texture place it. An illustration made in Voyager; the values on screen are the recipes below.

Walkie-talkie in a canyon: Base, this is Ridge Two. We found the trail marker, but the creek is too high to cross. We're turning back before dark. Through a wall at a party: I'm not hiding in here. I just needed five minutes. Tell him I'll come out when the song changes. Worn answering-machine tape: Hi love, it's Mum. It's Sunday, about six. Nothing important, just ring me back when you're home.

The recipes work in Premiere Pro, DaVinci Resolve, Audacity or any DAW. The last section shows how to make and keep them as styles in Voyager. For the built-in effects (telephone, cafe, old radio, cassette, vinyl and two sci-fi voices), see [make a voice sound like a phone call or old radio](/blog/phone-radio-voice-effects).

## Split the description into three parts

Read the description for the device, the place and the age. Each one maps to a few ordinary effects.

| The description says               | Part   | Reach for                                                                                                                             |
| ---------------------------------- | ------ | ------------------------------------------------------------------------------------------------------------------------------------- |
| walkie-talkie, intercom, comms     | Device | A narrow band (about 350 Hz to 3.2 kHz), a lift around 1.8 kHz, hard compression around 6:1, some drive, a bitcrusher at 8 to 10 bits |
| through a wall, next room, muffled | Device | Cut above 2.5 to 3 kHz and below about 120 Hz, dip 1.8 kHz by 3 to 6 dB                                                               |
| canyon, cave, hall, church         | Place  | A long reverb, 1.2 to 1.5 seconds, its tail rolled off around 2.5 to 4 kHz                                                            |
| booth, bathroom, small room        | Place  | A short reverb, 0.2 to 0.4 seconds                                                                                                    |
| at a party, in a crowd             | Place  | A murmuring crowd bed and a few glass clinks                                                                                          |
| worn tape, old machine             | Age    | Slow and fast pitch wobble, gentle compression around 2.5:1, high hiss                                                                |
| gritty, broken speaker             | Age    | More drive and fewer bits                                                                                                             |

Two or three rows usually carry a description. Add more only when you can hear what is missing.

The noise bed is the quietest part and does a lot of the work. Static, a crowd or hiss fills the gaps between words, which is where a listener decides whether the place is real. Keep it 20 dB or more under the voice.

## Decide where the listener stands

"A walkie-talkie in a canyon" is two different scenes. If you are in the canyon holding the radio, the canyon echoes what comes out of its speaker: radio first, then room. If the ranger is in the canyon and you are back at base, the echo travels through the radio with the voice: room first, then radio. The settings are the same. Only the order changes, and once the take is made, no other setting moves you from one scene to the other.

[Audio: Radio, then canyon: you are in the canyon](https://cdn.moda.app/voyager/blog/custom-voice-effects/walkie-canyon-92192c0d512685ef.mp3)

A man on a walkie-talkie: Base, this is Ridge Two. We found the trail marker, but the creek is too high to cross. We're turning back before dark.

[Audio: Canyon, then radio: you are back at base](https://cdn.moda.app/voyager/blog/custom-voice-effects/canyon-room-then-device-fc63ad24af1463d2.mp3)

The same line, with the canyon echo passing through the radio: Base, this is Ridge Two. We found the trail marker, but the creek is too high to cross. We're turning back before dark.

[Audio: The same line, clean](https://cdn.moda.app/voyager/blog/custom-voice-effects/clean-canyon-34782d93b4b8d633.mp3)

With the room second, the echo is ordinary reverb around a narrow voice. With the room first, the radio's band limit and compressor squeeze the echo along with the voice, so it is louder in the pauses and loses its top end. A noise bed works the same way: added after the device, it is on the listener's side; added before, it comes through the device.

## Three recipes

Each recipe lists the effects in the order to apply them.

### Walkie-talkie in a canyon

You are on the ledge with the radio: a search team, a film crew, a guard in a game. For a radio in a forest or a field, drop the room. For the base-camp version, move step 4 to the top.

1. High-pass at 350 Hz and low-pass at 3.2 kHz.
2. Boost 1.8 kHz by 4 dB.
3. Compress at 6:1, add about 6 dB of drive, and crush to 9 bits.
4. Add a 1.2 second reverb with its tail rolled off above 3 kHz.
5. Under it, white noise between 600 Hz and 6 kHz, 38 dB under the voice.

### Through a wall at a party

You are at the party, and the voice is in the next room.

1. High-pass at 120 Hz and low-pass at 2.8 kHz. A wall passes the low end and stops the top.
2. Lift 180 Hz by 2 dB and dip 1.8 kHz by 3 dB.
3. Add a 0.6 second room with its tail rolled off above 2.5 kHz.
4. Under it, a murmuring crowd about 22 dB down and a few glass clinks about 30 dB down.

[Audio: Through a wall at a party](https://cdn.moda.app/voyager/blog/custom-voice-effects/party-wall-3b03fe25e006924a.mp3)

A woman in the next room, through a wall, with a party murmuring around the listener: I'm not hiding in here. I just needed five minutes. Tell him I'll come out when the song changes.

[Audio: The same line, clean](https://cdn.moda.app/voyager/blog/custom-voice-effects/clean-party-575b2e40ef0b5db5.mp3)

The crowd comes after the wall, so it is on your side, which suits someone hiding from a party. For a party overheard from a quiet hallway, put the crowd before the filters. With no crowd recording, chop, pitch and smear a recording of your own until no words survive.

### Worn answering-machine tape

An old message on a cheap machine: a voicemail from the past in a documentary, a found message in a game, a flashback.

1. Add wow and flutter: two very short modulated delays, one slow (under 1 Hz) and one fast (around 5 Hz).
2. High-pass at 90 Hz and low-pass at 6 kHz.
3. Lift 180 Hz by 3 dB and dip 1.8 kHz by 2 dB.
4. Compress gently at 2.5:1 and add about 3 dB of drive.
5. Under it, white hiss between 3 and 12 kHz, 32 dB under the voice.

[Audio: Worn answering-machine tape](https://cdn.moda.app/voyager/blog/custom-voice-effects/answering-machine-7644d6372c944fcb.mp3)

An older woman on an old answering-machine tape: Hi love, it's Mum. It's Sunday, about six. Nothing important, just ring me back when you're home.

[Audio: The same line, clean](https://cdn.moda.app/voyager/blog/custom-voice-effects/clean-machine-fc14389bdf4fb8f3.mp3)

## Change one thing, then listen again

Change one setting per listen. Change two and you can't tell which one you heard. On the canyon we wanted a bigger echo, so we changed only the reverb: 1.4 seconds instead of 1.2, and 4 dB louder.

[Audio: Canyon, one change: 1.4 s room, 4 dB louder](https://cdn.moda.app/voyager/blog/custom-voice-effects/walkie-canyon-more-echo-8cd712fd8048e931.mp3)

| You want         | Change                                              |
| ---------------- | --------------------------------------------------- |
| More echo        | Reverb 0.2 seconds longer and 4 dB louder           |
| Darker           | Low-pass about a quarter lower (3.2 kHz to 2.5 kHz) |
| Brighter         | Low-pass about a third higher                       |
| More static      | Noise bed 3 to 4 dB louder                          |
| Grittier         | One bit fewer on the crusher, or 2 dB more drive    |
| Easier to follow | Quieter beds and room, less drive, a wider band     |

## Make and keep a style in Voyager

In Voyager you describe the sound and the agent does the translation in the first table. With a finished clean take in the Voice panel, ask:

```text
Make this take sound like a walkie-talkie in a canyon. Play me a preview before you make anything.
```

The agent answers with a style card that previews the first 5 seconds of your take, with **Clean** and **Styled** players. Each **Change one thing** chip, such as **More echo** or **Darker**, makes one change and plays a new preview: the rule above, as a button. **Adjust manually** opens the same settings as sliders. When it sounds right, choose **Save to My styles**, then **Make styled take**.

![The three illustrated scenes again, each starting on its finished style with a row of chips under the slider panel: More echo, Less echo, Darker, Brighter, More noise, Cleaner. On the canyon, Less echo takes Room size from 1.2 to 1.0 seconds and Room mix from +14 to +10 dB. On the party, More noise adds a hiss at −36 dB. On the answering machine, Brighter raises High cut from 6,000 to 8,000 Hz. Only the sliders the chip changes move.](https://cdn.moda.app/voyager/blog/custom-voice-effects/one-change-poster-4847c7605b8ec260.webp)

[Video: The three illustrated scenes again, each starting on its finished style with a row of chips under the slider panel: More echo, Less echo, Darker, Brighter, More noise, Cleaner. On the canyon, Less echo takes Room size from 1.2 to 1.0 seconds and Room mix from +14 to +10 dB. On the party, More noise adds a hiss at −36 dB. On the answering machine, Brighter raises High cut from 6,000 to 8,000 Hz. Only the sliders the chip changes move.](https://cdn.moda.app/voyager/blog/custom-voice-effects/custom-styles-one-change-867a36aba849b121.mp4)

One tap, one change, then listen again. An illustration made in Voyager; the values on screen are what each chip changes in the app.

Walkie-talkie in a canyon, before and after Less echo: Base, this is Ridge Two. We found the trail marker. Through a wall at a party, before and after More noise: I'm not hiding in here. I just needed five minutes. Worn answering-machine tape, before and after Brighter: Hi love, it's Mum. It's Sunday, about six.

Saved styles appear under **My styles** in the Style picker, in every project. Each styled take keeps its own copy of the recipe, so editing a style later doesn't change takes you already made. Styling runs on your computer and uses no credits. Previews and saving are free; a full styled take needs a paid plan, Starter and up.

Three limits. A custom style always puts the room after the device, so Voyager makes the in-the-canyon reading, not the base-camp one, and a crowd is always on your side of the wall; for the other reading, use your editor. There is no pitch shift or ring modulation, so for a robot or masked voice use the built-in Trooper and Masked host styles. And a styled take ends half a second after the last word, so leave silence at the end of the clean take if a long room has to ring out.

## If it still sounds fake

Listen to the gaps between words, not the words. If the pauses are dead, the place isn't there yet: add or raise the bed. If the echo in the pauses puts the listener in the wrong spot, swap the order before you touch another setting.

Style your own voice, a stock voice or one you have permission to use. Label reenactments as reenactments, and don't make anything that passes as a real person's message.


## Voyager is a desktop AI agent for creative work

Describe what you want to make; Voyager plans it, uses your apps and tools, and keeps everything editable.

[Sign up](/sign-up)