← Discover
Discover · GitHub · VoxCPM↑ +658 today

Tokenless Multilingual Sound Design

VoxCPM; It is a tokenizer-free open source TTS model developed for multilingual speech production, creative sound design and realistic voice cloning.

Updates

  • September 2, 2026: Stars 34,774 → 36,548, latest release 2.0.3 (May 11, 2026).
  • August 2, 2026: Stars 22,648 → 34,774, latest release 2.0.3 (May 11, 2026).

Installation

Install via PyPI
pip install voxcpm

Running it

Generate speech from text
voxcpm design --text "VoxCPM2 brings studio-quality multilingual speech synthesis." --output out.wav

What does it offer?

  • Multilingual and natural speech production.
  • Creative and original sound design.
  • High accuracy voice cloning.

Responsibility note

Includes voice cloning. Impersonating a person's voice without permission creates legal and ethical problems; The responsibility belongs to the user.

How to install, how to use?

🤖 Paste this into your AI agent (Claude Code · Codex · Antigravity)

Run 'pip install voxcpm' to install the VoxCPM text-to-speech tool, then generate an audio file with the command 'voxcpm design --text "Hello world" --output out.wav'; If I want, clone that sound using 'voxcpm clone' with a reference sound file.

Related dictionary terms

Tokenizer-freeVoice CloningCloningText-to-SpeechCloneOpen Source
Who it is forVoice/TTS developers, researchers
DifficultyAdvanced · ML/model knowledge
What offersMultilingual TTS + sound design + cloning
FeeFree · open source (Apache-2.0)
LicenseApache-2.0 · ayrıntı aşağıda
License: Apache-2.0 · code freely available/commercial. You assume legal/ethical responsibility for model output and audio transcription.

Links

TreScout did not build this tool · we found it in GitHub trends and wrote it up. This page describes the repository as of 2026-05-30: The star count and our text belong to that day, the repository may have changed since. Check the repository link for the current state. This page was machine-translated from the Turkish original · the Turkish version prevails.