https://github.com/varppi/ssvc
Somewhat Secure Voice Changer
https://github.com/varppi/ssvc
anonymity opsec privacy security voice-changer
Last synced: about 2 months ago
JSON representation
Somewhat Secure Voice Changer
- Host: GitHub
- URL: https://github.com/varppi/ssvc
- Owner: varppi
- License: gpl-3.0
- Created: 2025-02-12T06:39:07.000Z (8 months ago)
- Default Branch: main
- Last Pushed: 2025-03-08T15:23:43.000Z (7 months ago)
- Last Synced: 2025-08-09T08:51:22.355Z (about 2 months ago)
- Topics: anonymity, opsec, privacy, security, voice-changer
- Language: C#
- Homepage:
- Size: 37.1 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
### 
[](https://img.shields.io/badge/c%23-%23239120.svg?style=for-the-badge&logo=csharp&logoColor=white)
[](https://img.shields.io/badge/.NET-5C2D91?style=for-the-badge&logo=.net&logoColor=white)
### INSTALL FROM RELEASES
SSVC (Somewhat Secure Voice Changer) is a voice changer that is much harder to reverse and indetify people behind than traditional voice changers. It does this by randomly alternating the pitch and/or different frequencies in the sound.
## Why?
If you ever need to talk to someone that you suspect might try figure out information about you for malicious purposes, you would probably want to alter your voice so that it cannot be used to identify you. Traditional voice changers typically do the job, but with the advancements of AI, I was able to correlate the pitch shifted version of a person's voice so it was clear that a normal voice changer would be insufficient to conceal your identity if the attacker had another recording of you **WITHOUT** the voice changer.
## How?
To combat speaker recognition AIs, I combined constantly shifting the pitch and randomly making certain frequencies louder or quiter, which worked on the AI flawlessly. You had to turn down the distortion factor a significant amount before it could detect that it was the same person again.
## Demo
I used `speechbrain/spkrec-ecapa-voxceleb` for speaker identification.
### With SSVC
**Settings:**
```
Pitch: -25
Distortion: 10
Voice volume: 10
- reduce background noise
```
**Detection:**
```
score: -0.0162
is same person: FALSE
```
### Without SSVC
**Detection:**
```
Score: 0.8336 (VERY SIMILAR)
is same person: TRUE
```
## Screenshots
