Voice Cloning
Clone a voice on your Windows PC
Choose a reference clip, pick a compatible model, and generate new speech in that voice. The clip and the generated audio stay on your computer.
Free plan with Kokoro voices, scripts up to 1,500 characters · 3-day Pro trial, then $49 once
voice cloning software Windows
Murmur runs locally on 64-bit Windows 10 and 11, so your text stays on your PC and you can export WAV without cloud APIs or per-word fees.
- Reference cloning from a short clip
- Local generation on Windows
- Several compatible models
- WAV export
01 · Context
How voice cloning works in Murmur
Voice cloning builds a synthetic voice from a recording of a real speaker. You provide a clean reference clip, the model uses it to condition the new speech, and then you type the words you want that voice to say.
On Windows, cloning is available through Chatterbox Turbo, Chatterbox Multilingual, Qwen3-TTS Base, SparkTTS, Breeze TTS 2, and OmniVoice. Some models also ask for a transcript of the reference clip so the model can separate the voice from the words.
02 · Details
Which Windows model should you use?
| Model | Good for | Notes |
|---|---|---|
| Chatterbox Turbo | Fast English cloning | Also supports paralinguistic tags |
| Chatterbox Multilingual | Cloning across 23 languages | Choose the language before you generate |
| Qwen3-TTS Base | Short reference clips | Needs a transcript of the reference clip |
| SparkTTS | Pitch and rate control | Needs a transcript of the reference clip |
| Breeze TTS 2 | English and Chinese, style direction | The heaviest model, and a GPU helps most here |
| OmniVoice | Broad language coverage | Needs a transcript of the reference clip |
03 · Context
What makes a good reference clip
Cloning quality follows sample quality. Use one speaker, a quiet room, a steady distance from the microphone, and a natural reading pace. Ten clean seconds can be enough. Two noisy minutes are not better.
Avoid background music, overlapping speakers, heavy processing, and clipping. If the clip sounds clear on headphones, it is usually good enough to build a usable voice.
04 · Workflow
A simple cloning workflow on Windows
05 · Context
Consent and licensing
Only clone a voice you have permission to use. A cloned voice is tied to a real person, and impersonation or misleading audio can be restricted or illegal. Get consent before you clone someone else, and follow the model licence for commercial work.
Murmur's purchase licenses the app, not the voice. You are responsible for the rights to the script, the reference recording, and the way you use the generated audio.
06 · Context
What the Windows beta does not include
The Windows beta does not have the Mac app's projects, batch queue, community voice browsing, CLI or MCP automation, MDM deployment, M4A export, or automatic updates. It is a single-script cloning workflow with WAV export.
Fish S2 Pro is marked Coming soon on Windows. The other cloning models listed above work in the beta.
FAQ
Common questions
Yes. Use a compatible Pro model, choose a clean reference clip, and generate new speech locally. Some models also ask for a transcript of the clip.
Windows · Local TTS
Build a reusable voice on your PC
Download Murmur for Windows, try a cloning model during the 3-day Pro trial, and keep Pro for $49 once.