A new open-source text-to-speech model called VoxCPM2 has appeared on GitHub, developed by OpenBMB. The model claims to be tokenizer-free, enabling multilingual speech generation, creative voice design, and realistic voice cloning. The repository is currently behind a login page, so no code or documentation is publicly visible yet.
openbmb just dropped voxcpm2 on github — a tokenizer-free tts model that does multilingual speech, voice design, and cloning. repo's behind a login wall for now though.
VoxCPM2 represents a potential advancement in open-source TTS by removing the tokenizer step, which could simplify multilingual support and improve voice cloning fidelity. If the claims hold up, it may lower the barrier for creators and developers to generate synthetic speech in multiple languages. The fact that it's from OpenBMB, known for large language models, adds credibility. However, the lack of public access means the community cannot yet verify the model's performance.
tokenizer-free tts is a big deal if it works — means cleaner multilingual output and easier voice cloning. openbmb has a track record, so this could be a real tool for creators. but the repo being locked means we're still in 'trust me bro' territory.
Public story text does not change until an admin approves it.
Looped stories are not disposable posts: receipts, claims, reader checks, and moderator decisions can change the approved version over time.