Xyron Voice Changer

Features

Built like a studio plug-in

Most voice changers just speed up or slow down your voice. Xyron Voice Changer separates pitch (how high you speak) from formant (the size of your throat) so a deep voice sounds deep and a female voice sounds like a person, not a chipmunk.

Formant-aware pitch shifting

Phase-locked Laroche–Dolson peak shifting plus a cepstral spectral envelope keep pitch and timbre independent, with far less of the “phasey” robotic artefact.

37 hand-tuned presets

Female, Male, Character and Effect voices: from Natural Female and Radio DJ to Demon, Ghost, Walkie-Talkie and Underwater. Save your own in the Mine tab.

Studio cleanup

Spectral noise suppression with a Learn room noise button, noise gate, de-esser, 4-band EQ, compressor and a look-ahead peak limiter that never clips.

XY voice pad

Drag a single dot to move pitch and formant together: up for higher, down for deeper, right for a more feminine timbre. Double-click to snap back to your natural voice.

Character & FX rack

Robot vocoder, whisper, breath, vibrato, octave harmonizer, distortion, bitcrusher, chorus, phaser, echo and reverb, all adjustable live while you talk.

Global hotkeys

Ctrl+Alt+B raw voice, M mute, V on/off. They work even while a full-screen game or Discord has focus.

Hear what others hear

Record 5 s and listen plays back exactly what your friends receive, and Hear myself monitors live through your headphones. Click-free A/B with Raw voice.

Offline file conversion

Run any WAV, MP3, FLAC, OGG, M4A or AIFF file through the same engine and save a 24-bit WAV, from the app or from the command line.

Private by design

No network code at all: no account, no cloud, no telemetry. Your voice never leaves your PC. Settings live in a plain JSON file you can read.

English & Turkish

The whole interface, every preset and the CLI are translated. The language follows your Windows display language on first launch and can be switched any time.

Self-healing audio

Adaptive micro-resampling absorbs clock drift between sound cards, a watchdog restarts dropped streams, and live latency, CPU and buffer health are always on screen.

Open & tested

Clean, typed Python with 113 automated tests covering the DSP engine, effects, audio routing, translations, presets and CLI.

Main window with the Studio Male preset selected — Main window: device routing, live telemetry, preset browser and the voice-identity pad.

Character category with the Demon preset selected — Character voices: selecting *Demon* moves the XY pad to −6 st pitch / −3.5 st formant.

Under the hood

How your voice travels

Audio is processed in mono at 48 kHz. A dedicated audio thread does the heavy lifting while the interface only updates parameters, so dragging a slider never interrupts your voice.

Input

Microphone

WASAPI · MME · DirectSound
Thread-safe ring buffer

Spectral engine (STFT)

Voice identity

1 · Noise suppression
2 · Pitch shift (phase-locked)
3 · Formant via cepstral envelope
4 · Robot · whisper · breath · harmonizer
5 · De-esser

Pedalboard chain

Tone & character

Gate → Low/High cut
Bass · Body · Clarity · Brightness
Distortion · Bitcrush
Chorus · Phaser
Compressor → Echo → Reverb

Safety

Limiter

Look-ahead peak limiter
tanh soft ceiling (−0.2 dBFS)
Click-free bypass & mute ramps

Output

VB-CABLE + headphones

Discord, OBS, games, calls
Optional self-monitoring
Adaptive ±0.5% resampling

Effects are only instantiated when they are switched on, and the chain is rebuilt only when its structure changes, so reverb and echo tails keep ringing while you move sliders.

Get started

Up and running in 5 minutes

You need Windows 10 or 11 (64-bit), a microphone and, to send your new voice to other apps, the free VB-CABLE virtual audio driver.

Install the virtual microphone (one time only)

Download VB-CABLE from the official VB-Audio website (or the unmodified copy attached to our release), extract it, then right-click VBCABLE_Setup_x64.exe → Run as administrator → Install Driver, and restart your PC.

Tip: extract it into a folder named VBCable next to the app and the in-app “VB-Cable not found: Install” button will launch the installer for you.

Download and launch Xyron Voice Changer

Grab Xyron-Voice-Changer-win64.exe from the latest release and double-click it. No installation needed. The first launch takes 5–10 seconds while it unpacks.

“Windows protected your PC”: click More info → Run anyway. The app is not code-signed yet, so SmartScreen shows this for every new build.
Antivirus says “Access denied”: add the EXE or its folder to your antivirus exclusions, or use the portable folder build, which starts faster and triggers fewer false positives.

Pick your devices

In the left sidebar under Audio devices:

Audio system: Windows WASAPI (recommended; try MME if you hear crackles)
Microphone: your real microphone
Virtual microphone: CABLE Input (VB-Audio Virtual Cable)
Headphones: optional, to hear yourself

Press START and choose a voice

Click START, then Learn room noise and stay silent for 2–3 seconds. Click any voice card. Use Record 5 s and listen to hear exactly what others will hear.

Point your apps at the cable

In Discord, OBS, your game or meeting app, set the input device / microphone to CABLE Output (VB-Audio Virtual Cable). See the app-specific guides below.

App-specific setup

Open User Settings → Voice & Video.
Set Input Device to CABLE Output (VB-Audio Virtual Cable).
Turn off Noise Suppression (Krisp) and Echo Cancellation: Xyron Voice Changer already cleans your voice, and Krisp can mangle shifted voices.
Set Input Sensitivity to automatic or use Push-to-Talk as usual.
Use Let's Check in Discord, or Xyron's Record 5 s and listen, to verify.

In Sources, click + → Audio Input Capture.
Choose Device: CABLE Output (VB-Audio Virtual Cable).
Mute or remove your original Mic/Aux source so viewers don't hear both voices.
Optional: use the global hotkey Ctrl+Alt+B on stream to toggle between your real and changed voice.

Open the game's Audio / Voice chat settings.
Set the Microphone / Input device to CABLE Output (VB-Audio Virtual Cable).
If the game has no device selector it uses the Windows default: in Sound settings → Input, set CABLE Output as the default device.
Global hotkeys keep working while the game is full-screen.

Open Settings → Audio in Zoom, Teams or Google Meet.
Choose CABLE Output (VB-Audio Virtual Cable) as the Microphone.
Disable the app's own heavy noise suppression for the cleanest result.
For professional calls, try the Studio Male preset: your own voice, just cleaner.

Reference

Every control, explained

All 31 parameters update in real time. st = semitones (12 st = one octave). Percent values run from 0 (off) to 100%.

Section	Control	Range	Default	What it does

Engine quality modes

Choose in Audio devices → Engine quality. Engine latency is the STFT window; your sound card adds a few more milliseconds.

Mode	FFT / hop	Engine latency	Best for
Maximum Clarity max	4096 / 512	≈ 85 ms	Deep voices that sound grainy when shifted
Studio Quality (default) studio	2048 / 256	≈ 43 ms	Everyday use, streaming, voice chat
Balanced balanced	2048 / 512	≈ 43 ms	Older CPUs (half the processing)
Low Latency low_latency	1024 / 256	≈ 21 ms	Competitive games, singing along

Global hotkeys

Registered system-wide through the Windows API, so they work while any other app has focus. If another program already owns a combination, the app tells you on start-up.

Shortcut	Action
`Ctrl`+`Alt`+`B`	Raw voice: instantly bypass the changer (click-free A/B)
`Ctrl`+`Alt`+`M`	Mute the microphone
`Ctrl`+`Alt`+`V`	Start / stop the voice changer

💡Hotkeys can be turned off with "hotkey_enabled": false in settings.json.

Command line

Scriptable, too

Run main.py without arguments to open the app. With arguments, it becomes a command-line tool for batch conversion and diagnostics. Output follows --lang, your saved language, or your system language.

List all presets

python main.py --list-presets --lang en

Output (excerpt)

f_natural        Female    Natural Female       pitch=+10.0 formant=+3.2
m_deep           Male      Deep Male            pitch=-3.0 formant=-1.5
c_robot          Character Robot                pitch=+0.0 formant=+0.0
e_radio          Effect    Walkie-Talkie        pitch=+0.0 formant=+0.0

Show audio devices

python main.py --list-devices

Output (example)

== Windows WASAPI
  IN  [12] Microphone (USB Audio)  <- suggested microphone
  OUT [ 9] CABLE Input (VB-Audio Virtual Cable)  <- virtual cable
  OUT [ 7] Headphones (Realtek Audio)  <- headphones

Convert a file with a preset

python main.py --render podcast.mp3 podcast_trailer.wav --preset m_trailer

Custom pitch and formant

# +5 semitones higher, throat 2 semitones smaller
python main.py --render in.wav out.wav --pitch 5 --formant 2 --quality max

Batch-convert a folder (PowerShell)

Get-ChildItem .\voice-lines\*.wav | ForEach-Object {
  python main.py --render $_.FullName ".\out\$($_.BaseName)_robot.wav" --preset c_robot
}

Option	Description
--list-presets	List built-in and saved presets
--list-devices	List audio devices per audio system, with suggestions
--render IN OUT	Process WAV/MP3/FLAC/OGG/M4A/AIFF → 24-bit 48 kHz WAV
--preset KEY	Preset key (default `m_clean`)
--pitch / --formant N	Override pitch / formant in semitones (−12…+12)
--quality MODE	`max` · `studio` · `balanced` · `low_latency`
--lang tr\|en	Language for CLI output (and the app window)
--version / --debug	Show version · show full traceback on errors

Exit codes: 0 success · 1 runtime error · 2 usage error (missing file, unknown preset).

For developers

Run, test and build from source

Requires Python 3.12+ on Windows. Dependencies: numpy, sounddevice (PortAudio), pedalboard, customtkinter and Pillow.

1 · Clone and install

git clone https://github.com/Xyronb1/xyron-voice-changer.git
cd xyron-voice-changer
python -m venv .venv
.venv\Scripts\python.exe -m pip install -r requirements.txt

2 · Run the app

.venv\Scripts\python.exe main.py

3 · Run the test suite

.venv\Scripts\python.exe -m pip install pytest
.venv\Scripts\python.exe -m pytest

4 · Build the EXE (PyInstaller)

.venv\Scripts\python.exe -m pip install pyinstaller
# single portable file -> dist\Xyron Voice Changer.exe
.venv\Scripts\python.exe -m PyInstaller --noconfirm --clean Xyron_Voice_Changer.spec
# folder build: faster start, fewer antivirus false positives
.venv\Scripts\python.exe -m PyInstaller --noconfirm --clean Xyron_Voice_Changer_onedir.spec

Project layout

main.py              entry point: GUI + CLI
xyrox_voice/
  dsp.py             STFT engine: denoise, pitch, formant, layers
  effects.py         Pedalboard chain + look-ahead limiter
  processor.py       engine + effects + meters, quality modes
  audio_io.py        devices, ring buffers, real-time router
  ui.py              customtkinter interface
  presets.py         37 presets + user preset store
  params.py          all parameters, ranges, formatting
  i18n.py            English / Turkish translations
  hotkeys.py         global Windows hotkeys (ctypes)
  render.py          offline file rendering
  config.py          settings.json
tests/               pytest suite

🌍Add a language: add its code to LANGUAGES in xyrox_voice/i18n.py and a dictionary to TRANSLATIONS. Missing strings fall back gracefully, and tests/test_i18n.py tells you exactly which ones are untranslated.

Help

FAQ & troubleshooting

My friends can't hear my changed voice.

Check the chain end to end: in Xyron, Virtual microphone must be CABLE Input and the app must be started (green “Active”). In Discord or your game, the input device must be CABLE Output. If VB-CABLE doesn't appear anywhere, reinstall it as administrator and reboot.

I hear crackling or dropouts.

Switch Audio system to MME or Windows DirectSound.
Select Engine quality → Balanced to halve the CPU load.
Close other apps that use the microphone in “exclusive mode”.
Watch the BUFFER indicator: Good / Fair / Poor tells you how healthy the stream is.

There is too much delay.

Choose Engine quality → Low Latency (≈21 ms engine latency) and use WASAPI. Note that Hear myself always sounds slightly delayed and “doubled” because you also hear your real voice through your skull; Record 5 s and listen gives a more honest impression.

The changed voice sounds grainy or metallic.

Very large pitch shifts on deep voices are the hardest case. Select Engine quality → Maximum Clarity (about 40 ms more latency, much cleaner). Also press Learn room noise while silent, and avoid stacking Discord's Krisp on top.

Windows SmartScreen or my antivirus blocks the EXE.

The executable is built with PyInstaller and is not code-signed yet, which some antivirus engines flag by reputation alone. Choose More info → Run anyway, add an exclusion, use the portable folder build, or run from source. Every release lists SHA-256 checksums so you can verify the download.

Where are my settings and saved profiles?

In %APPDATA%\Xyron_Voice_Changer\: settings.json (devices, language, last preset) and presets.json (your Mine profiles). Delete the folder to reset everything. Settings from the earlier XYROX Voice Studio builds are migrated automatically.

How do I change the language?

Use the Dil / Language selector at the bottom of the left sidebar. On first launch the app picks Turkish on a Turkish Windows and English everywhere else.

Does it work on macOS or Linux?

The DSP engine, effects and CLI are cross-platform Python, but the app targets Windows: global hotkeys use the Win32 API and the virtual cable workflow is built around VB-CABLE. Ports are welcome; see the developer section.

Is it really free? Is my voice recorded?

Yes, it's free and MIT-licensed. Nothing is recorded or uploaded: the source contains no networking code. The 5-second test recording stays in memory and is discarded after playback.

The team

Developers

Xyron Voice Changer is built by a small human + AI team.

Xyron

Creator · Lead developer

Project owner. Product vision, voice design and preset tuning, testing on real hardware, and releases.

@Xyronb1

Claude

AI co-developer · by Anthropic

Xyron's AI engineering partner. DSP engine and audio routing, internationalisation, the test suite, code review and this documentation.

@claude · anthropic.com/claude

mirac221

Developer

Core team member contributing development, testing and feedback.

@mirac221

Contributing

Bug reports, new presets, translations and pull requests are welcome. Please run python -m pytest before opening a PR and describe how you tested audio changes. Open an issue to start a discussion.

Built with

Pedalboard (Spotify) for effects and audio I/O · python-sounddevice / PortAudio for real-time streams · CustomTkinter for the interface · NumPy for the DSP · VB-CABLE by VB-Audio (donationware) for the virtual microphone.

⚖️Use responsibly. Voice changers are for fun, privacy and creativity. Don't use Xyron Voice Changer to impersonate real people, deceive, harass or break the rules of the platforms you use.

One sentence, twelve voices

Built like a studio plug-in

Formant-aware pitch shifting

37 hand-tuned presets

Studio cleanup

XY voice pad

Character & FX rack

Global hotkeys

Hear what others hear

Offline file conversion

Private by design

English & Turkish

Self-healing audio

Open & tested

How your voice travels

Microphone

Voice identity

Tone & character

Limiter

VB-CABLE + headphones

Up and running in 5 minutes

Install the virtual microphone (one time only)

Download and launch Xyron Voice Changer

Pick your devices

Press START and choose a voice

Point your apps at the cable

App-specific setup

37 built-in voices

Every control, explained

Engine quality modes

Global hotkeys

Scriptable, too

Run, test and build from source

FAQ & troubleshooting

Developers

Xyron

Claude

mirac221

Contributing

Built with