VoxCPM vs csm
Speech generation projects compared
VoxCPM and csm are open-source Python projects that generate speech from text. VoxCPM focuses on multilingual synthesis, voice design, and cloning, while csm generates speech with optional conversational audio and transcript context.
VoxCPM: Generate and Clone Multilingual Speech
VoxCPM is a tokenizer-free text-to-speech system for multilingual speech generation, voice design, and voice cloning. Its VoxCPM2 release targets teams and developers who need expressive speech synthesis and can support a 2B-parameter model.
csm: Generate Conversational Speech from Text and Audio
CSM is Sesame’s speech-generation model, producing audio from text and optional conversation context. It suits developers building voice experiences who can run large models on a CUDA-compatible GPU and provide the required Hugging Face checkpoints.
| VoxCPM | csm | |
|---|---|---|
| Language | Python | Python |
| License | Apache-2.0 | Apache-2.0 |
| Stars | 38.3k | 14.7k |
| Forks | 4.3k | 1.5k |
| Last analyzed | Oct 3, 2026 | Oct 3, 2026 |
Key differences
- VoxCPM supports 30 languages and natural-language voice design; csm focuses on generating speech from text with selected speaker identities and conversation context.
- VoxCPM offers reference-audio voice cloning, style control, streaming generation, and fine-tuning; csm conditions generation on prior transcript, speaker, and audio segments.
- VoxCPM uses a tokenizer-free diffusion autoregressive approach; csm combines a Llama backbone with a Mimi audio decoder.
- VoxCPM documents Python and CLI workflows, serving integrations, and edge-inference options; csm provides Python examples and a quickstart, with checkpoints hosted separately on Hugging Face.
- VoxCPM documents about 8 GB of VRAM for VoxCPM2; csm requires a CUDA-compatible GPU and access to CSM-1B and Llama-3.2-1B checkpoints.
- Both projects use the Apache-2.0 license. VoxCPM has 38.3k stars and 4.3k forks, compared with csm at 14.7k stars and 1.5k forks.
Choose VoxCPM if you…
- need multilingual speech generation, including support for 30 languages.
- want natural-language voice design or reference-audio cloning with optional style guidance.
- need streaming or fine-tuning options and can run a substantial model.
Choose csm if you…
- want speech generation conditioned on conversational audio and transcripts.
- need to generate dialogue with multiple speaker identities from application-provided text.
- plan to pair speech output with a separate language model that generates text.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.