This 2B Model Cloned My Voice Locally (VoxCPM)

VoxCPM2 is a new open-source local text-to-speech model from OpenBMB that can generate natural 48 kHz speech, design entirely new voices from text, clone voices from short reference clips, and support around 30 languages without sending your audio or text to a hosted API. In this video, I install VoxCPM2, run it locally, generate speech from Python, serve it behind an OpenAI-style `/v1/audio/speech` endpoint with vLLM, test voice design from a simple prompt, clone my own voice, and try multilingual generation to see how practical it is for real developer workflows. 🔗 Relevant Links VoxCPM2 Repo - https://github.com/OpenBMB/VoxCPM HuggingFace Demo - https://huggingface.co/spaces/openbmb/VoxCPM-Demo VoxCPM - https://voxcpm.com/en/ ❤️ More about us Radically better observability stack: https://betterstack.com/ Written tutorials: https://betterstack.com/community/ Example projects: https://github.com/BetterStackHQ 📱 Socials Twitter: https://twitter.com/betterstackhq Instagram: https://www.instagram.com/betterstackhq/ TikTok: https://www.tiktok.com/@betterstack LinkedIn: https://www.linkedin.com/company/betterstack 📌 Chapters: 0:00 The Open Source Alternative to Eleven Labs 0:33 What Is VoxCPM2? 2:06 Installing VoxCPM2 Locally 3:15 Local Voice Generation 4:24 Serving VoxCPM2 With vLLM 5:35 AI Voice Design From Text 6:22 Local Voice Cloning Demo 7:00 Multilingual Voice Cloning 7:33 VoxCPM2 vs ElevenLabs, Qwen3-TTS & Chatterbox 8:15 VoxCPM2 Pros, Cons & VRAM Requirements 9:20 Should Developers Use VoxCPM2? 9:55 Final Verdict: Is VoxCPM2 Worth It?