Voice Cloning — Clone a Voice for Your Videos, On-Device
Clone a voice from a short reference clip and narrate your videos in it — running 100% on your own machine with the open-source VoxCPM engine, so your reference audio never leaves your device. No cloud upload, no subscription.
Voice Cloning at a glance
{
"tool": "ViralMint Voice Cloning",
"input": "10-30s reference clip (record or upload) + transcript",
"transcript": "auto-filled locally with Whisper, or typed",
"engine": "VoxCPM (open-source on-device voice model)",
"runs": "100% local — reference audio never leaves the device",
"output": "narration in the cloned voice, mp3 → your video",
"use_in": ["Voice-over tool", "Smart Video", "Brainrot"],
"platforms": ["macOS (Apple Silicon)", "Linux"],
"price": "5c / 1K chars, capped at 20c per generation",
"subscription": false,
"watermark": false
}Runs in the open-source ViralMint desktop app. Compare with preset AI voices. See also Smart Video.
What is voice cloning in ViralMint?
Voice cloning learns a voice from a short reference clip — 10 to 30 seconds of one speaker — and then narrates your scripts in that voice. Clone your own voice so faceless videos still sound like you, or any voice you have permission to use. It's different from a preset neural voice: instead of picking from a fixed catalog, you bring the voice.
The defining property is that it's on-device. ViralMint uses the open-source VoxCPM model, installed once into the desktop app (~545 MB, like the Chromium download). Registering a voice and generating speech both run on your own machine — your reference audio and the output never touch a server. For a feature where the input is someone's voice, running locally is the privacy answer cloud cloning services can't give. It works on macOS (Apple Silicon) and Linux.
How to clone a voice in 5 steps
- Install the voice engine. Open Tools → Voice Cloning and install the one-time on-device engine when prompted — it's on-demand (~545 MB, like the Chromium install), not bundled into the base app.
- Add a reference clip. Record straight from your mic or upload a clean 10–30 second clip of a single speaker, and give the voice a name.
- Auto-transcribe. Leave the transcript blank and ViralMint transcribes the clip locally with Whisper, or type the exact words yourself for precision.
- Clone the voice. Click Clone voice — it's registered on your machine and added to your voice library. Your reference audio never leaves the device.
- Narrate in it. Pick the cloned voice in the Voice-over tool, Smart Video or Brainrot. Your script is narrated in the cloned voice and exported as part of the finished video.
What's in Voice Cloning
100% on-device
The VoxCPM engine runs locally. Your reference clip and the generated audio never leave your computer — the privacy answer cloud cloning can't offer.
Clone from a short clip
10–30 seconds of one speaker is enough. Record from your mic or upload a file; the transcript is auto-filled locally with Whisper.
Use it everywhere
A cloned voice shows up as a "Clone voice" option in the Voice-over tool, Smart Video and Brainrot — pick it like any other voice.
A few cents, capped
5¢ per 1,000 output characters, capped at 20¢ per generation, from prepaid credits — no subscription, no per-minute cloud bill.
Open-source engine
Powered by the Apache-2.0 VoxCPM model inside the open-source ViralMint app — auditable, no proprietary cloud lock-in.
Manage your voices
Keep a library of cloned voices — play each reference clip, delete the ones you don't need. Delete once and it's gone everywhere.
Frequently asked
What is ViralMint voice cloning?
A feature that learns a voice from a short reference clip (10–30s of one speaker) and narrates your videos in it. Clone your own voice for faceless content, or any voice you have permission to use — and it all runs on your own machine via the open-source VoxCPM engine.
Does it run locally or in the cloud?
100% locally. The VoxCPM engine installs as a one-time ~545 MB download into the desktop app; registering a voice and generating speech both happen on your computer, so your reference audio is never uploaded.
How much does it cost?
5¢ per 1,000 output characters, capped at 20¢ per generation, from prepaid credits — no subscription. The model runs locally so there's no per-minute cloud cost, and registered accounts get a small free daily allowance.
Which platforms are supported?
macOS (Apple Silicon) and Linux today. The Windows build isn't ready yet, so Windows shows a clear "not available on this platform" state — the rest of ViralMint's neural and Edge TTS voices still work there.
Where can I use a cloned voice?
Anywhere ViralMint picks a voice: the Voice-over tool, Smart Video and the Brainrot short-maker. It appears as a "Clone voice" option alongside the preset voices.
Clone your first voice
Voice Cloning ships in the open-source ViralMint desktop app for macOS and Linux. Free download, on-device, prepaid credits — no subscription, no watermark.