encrypt-my-voice

encrypt_voice.py is a Linux-friendly voice obfuscator for audio files. It does not do cryptographic audio encryption; instead, it derives a strong, repeatable distortion profile from SHA-256 and applies layered FFmpeg effects so speech becomes hard to recognize.

What it does

• computes SHA-256 from the input file and a seed
• cuts audio into many chunks
• changes pitch and timing per chunk
• narrows the voice band with filters
• adds vibrato, tremolo, bitcrushing, echo, and masking noise
• can optionally shuffle chunks for a much more destructive mode

Requirements

• Linux
• python3
• ffmpeg and ffprobe

Ubuntu/Debian:

sudo apt update
sudo apt install ffmpeg python3

Usage

Basic run:

python3 encrypt_voice.py input.wav -o output.wav

More aggressive:

python3 encrypt_voice.py input.wav -o output.wav --strength 5 --shuffle-chunks

Repeatable result with your own seed:

python3 encrypt_voice.py input.wav -o output.wav --seed "my-secret-passphrase"

Use one stable profile across different files:

python3 encrypt_voice.py clip1.wav -o clip1_masked.wav --seed "persona-42" --fixed-profile
python3 encrypt_voice.py clip2.wav -o clip2_masked.wav --seed "persona-42" --fixed-profile

Save processing metadata:

python3 encrypt_voice.py input.wav -o output.wav --meta output.json

Notes

• Default output is <input>.obfuscated.wav if -o/--output is omitted.
• If --seed is not provided, the script generates a random 256-bit seed and prints it after processing.
• The same input file + same seed + same settings produce the same output.
• --shuffle-chunks destroys intelligibility much more strongly, but it also destroys phrase order.
• This is best described as irreversible voice obfuscation, not real encryption.