In simple English
This tool turns written text into spoken voice files. It speaks thirty different languages and creates sounds in high quality. You can type a description of a desired voice to create it from scratch. You can also upload a short audio clip to clone any specific speaker. It needs a powerful computer graphics card to run locally and free of charge.
What it is
VoxCPM2 is an open-source, two-billion-parameter text-to-speech model built on a diffusion autoregressive architecture. It bypasses discrete tokenization to generate continuous audio latents directly, supporting streaming output and custom voice training.
What it is good for
Solo builders who need natural voiceovers, video dubbing, or multilingual narration without paying for external cloud speech services.
- Supports thirty languages without requiring language tags in the text.
- Creates entirely new voices using simple natural-language descriptions.
- Provides open weights under the Apache-2.0 licence for commercial use.
- Requires an NVIDIA graphics card with at least 8 GB of memory to run well.
- Voice design and controllable cloning results can vary and may need a few retries.
- Large model size needs significant disk space and download time.
Paid cloud text-to-speech APIs and rigid voice generation tools.
vLLM-Omni for high-throughput serving and ComfyUI for node-based workflows.
- fishaudio/fish-speech33k stars
Another open-source text-to-speech model supporting multilingual cloning and synthesis.
- snapcast/snapcast8.0k stars
A server and clients that play one audio stream in several rooms, kept in time with each other.
- MusicPlayerDaemon/MPD2.8k stars
A background music player with no interface of its own, controlled over a network by separate client programs.
- AcademySoftwareFoundation/OpenTimelineIO2.0k stars
An interchange format and library for editing timelines: the order and length of cuts, and references to the media, but not the media.
pip install voxcpm