Voice Replacer
shift.ai-ministries.com
- Free plan
- Windows
- Web
What Voice Replacer is
Voice Replacer changes the vocal identity in an audio or video clip while keeping its words, timing, inflection, and emotional delivery. It accepts MP3, MP4, WAV, OGG, and M4A files, with clips described as up to a few minutes. Users can choose among six included voices or provide a WAV or MP3 reference; five or more seconds of clean speech is described as ideal. The browser-based DSP version runs in a tab on any device and offers pitch and formant controls, EQ, tone, dynamics, reverb, presets, preview, and render-and-download. Its quality ceiling is lower than the neural version. The downloadable v0.1.1 neural bundle is for Windows 10/11 and needs a compatible NVIDIA GPU with at least 6 GB VRAM. It is configured for offline use after download, and saves WAV output in its app/outputs folder until deleted. The maker says audio is not uploaded and network calls are limited to an optional update check that can be disabled. The free tool requires no account.
Who it suits
Voice Replacer suits people who want to alter a clip’s vocal identity using a browser tool or a local Windows neural bundle. The neural version is for users with a compatible NVIDIA GPU; the browser version is available on any device.
Good to know, and what to watch
- Accepts five listed audio and video formats.
- Includes six target voices and reference-clip input.
- Browser version runs on any device.
- Downloaded neural bundle is configured for offline use.
- Free to use without an account.
- Browser version has a lower quality ceiling.
- Neural bundle requires Windows 10/11.
- Neural bundle needs an NVIDIA GPU with at least 6 GB VRAM.
- Processes one file at a time.
Plans
Every plan and priceFree
Free plan
The verdict
Voice Replacer offers a browser-based option and a downloadable neural bundle, with different system requirements and quality ceilings. Note the single-file limit and use voice references responsibly, as the maker advises against cloning without permission or using the tool to deceive or impersonate.
What the maker publishes
- Batch limit
- The v0.1 interface processes one file at a time; batch mode is listed as on the roadmap. source
- Browser controls
- The browser studio includes pitch and formant controls, EQ and tone settings, dynamics controls, reverb, presets, preview, and render and download. source
- Browser version
- A browser based DSP version runs in the tab on any device and provides slider based pitch and formant shifting, with a lower quality ceiling than the neural tool. source
- CLI
- The maker says the included inference CLI accepts a single file per call and can be scripted from PowerShell or bash. source
- Inputs
- It accepts audio or video files in MP3, MP4, WAV, OGG, and M4A formats, with clips described as up to a few minutes. source
- Neural version
- The downloadable v0.1.1 bundle is for Windows 10/11 and requires a compatible NVIDIA GPU with at least 6 GB VRAM. source
- Offline use
- The downloaded bundle includes its model weights and runtime and is configured for offline use after download. source
- Output
- The neural version saves WAV output in the bundle’s app/outputs/ folder until the user deletes it. source
- Privacy
- The maker says audio is never uploaded, inference is not hosted, and the tool makes no network calls except an optional update check that can be disabled. source
- Purpose
- Voice Replacer changes a clip’s vocal identity to a target voice while preserving its words, timing, inflection contour, and emotional delivery. source
- Responsible use
- The maker advises against cloning voices without permission or using the tool to deceive, harass, defame, or impersonate. source
- Support
- The product page provides troubleshooting guidance for CUDA errors, a busy port, antivirus quarantines, and out of memory errors. source
- Target voices
- Users can choose from six bundled voices or add a WAV or MP3 reference clip, with five or more seconds of clean speech described as ideal. source
Checked against the maker's pages on 7 Oct 2026.