Recorded in a kitchen, a car and a hall? Pick one voice, drop the clips in, and every clip comes out in that voice and that room. Same timing, frame by frame, so the lips still match.

Download for Mac Download for Windows

Version 4 · Mac (Apple Silicon, macOS 14+) and Windows 10/11 · free, offline, nothing leaves your computer

A laughing woman films herself from the driver's seat of a parked car, her phone on a small mount at the windscreen, a country road behind her

Requirements

Mac: Apple Silicon (M1 or newer), macOS 14+. Windows 10/11, 64-bit; an NVIDIA card makes it much faster. The installation downloads the voice models once (about 3 GB). After that it works offline. Nothing leaves your computer.

What it does to every clip

1 · CLEANRumble and steady noise out, gently.
2 · VOICESeed-VC turns the speech into your chosen voice. The length stays the same, frame by frame.
3 · PROFILEThe sound of your reference recording, its microphone and room, laid over the clip as an EQ curve.
4 · LEVELEvery clip at the same loudness (−16 LUFS), peaks held under −1 dB.
5 · EXPORTOne file per clip, or everything stitched into one. Video keeps its picture.

How to use it

  1. ① ADD FILES (click the number): any audio or video file (everything ffmpeg reads). Drag them into order; that order is also the stitch order.
  2. ② HOW TO EXPORT: ONE BY ONE, STITCHED or BOTH.
  3. ③ CONVERT (click the number). The first time it asks once for the voice recording every clip should sound like (10–30 seconds of clean, studio-sounding speech). The results open in a "Free Voice Convert" folder next to your first clip.

Nothing lost

Every file comes out in its own format: the same sample rate, the same bit depth and the same channels, as WAV (24-bit for MP3 or AAC sources). A video keeps its picture untouched, copied bit for bit, with the new sound as uncompressed PCM in a .mov (.mkv stays .mkv). Clips that match are stitched without re-encoding the picture.

Installing

Mac

Open the .dmg and drag the app to Applications. The first time, macOS asks: System Settings › Privacy & Security › Open Anyway (the app is signed by its author, not through Apple's paid program). At its first start it downloads the models itself.

Windows

Run the Setup. If SmartScreen warns, press More info › Run anyway. The installer opens a window that downloads Python, torch, Seed-VC and the models. When it closes, the app is ready.

Free and open

GPL-3.0. Voice conversion by Seed-VC (Plachtaa), vocoder BigVGAN (NVIDIA), speech features Whisper (OpenAI). No account, no upload, no subscription.

Put something in the bowl