# VoxCPM | OpenBMB | Knowledge Base | Instagit

VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning

GitHub Stars: 8.1k

Repository: https://github.com/OpenBMB/VoxCPM

---

## Articles

### [VoxCPM Ecosystem Tools: A Complete Guide to VoxCPM.cpp, ONNX, and ComfyUI](/OpenBMB/VoxCPM/what-voxcpm-ecosystem-tools-are-available-voxcpm-cpp-onnx-comfyui)

Explore the VoxCPM ecosystem with VoxCPM.cpp for fast inference, ONNX for easy deployment, and ComfyUI for node based TTS. Optimize your voice generation workflows today.

- Tags: how-to-guide
- Published: 2026-04-10

### [How to Run the VoxCPM Web Demo Locally: Complete Setup Guide](/OpenBMB/VoxCPM/how-to-run-the-voxcpm-web-demo-locally)

Learn how to run the VoxCPM web demo locally with this complete setup guide. Follow simple steps to clone the repository install the package and launch the demo interface for VoxCPM.

- Tags: how-to-guide
- Published: 2026-04-10

### [How to Deploy VoxCPM for High-Throughput Production with Concurrency](/OpenBMB/VoxCPM/how-to-deploy-voxcpm-for-high-throughput-production-with-concurrency)

Learn to deploy VoxCPM for high-throughput production with concurrency. Explore Nano-vLLM-VoxCPM for async batched inference or adjust Gradio settings for parallel processing.

- Tags: how-to-guide
- Published: 2026-04-10

### [CLI Commands for Batch Processing TTS Requests in VoxCPM: A Complete Guide](/OpenBMB/VoxCPM/what-are-the-cli-commands-for-batch-processing-tts-requests-in-voxcpm)

Master VoxCPM TTS batch processing with our guide. Learn the CLI commands to generate WAV files efficiently, supporting voice cloning and more. Accelerate your audio workflow today.

- Tags: how-to-guide
- Published: 2026-04-10

### [How to Use the VoxCPM Python API generate() Method for Text-to-Speech Synthesis](/OpenBMB/VoxCPM/how-to-use-the-voxcpm-python-api-generate-method)

Learn how to use the VoxCPM Python API generate() method for text to speech. Get a NumPy float32 waveform array for seamless audio synthesis. Explore the core functionality.

- Tags: how-to-guide
- Published: 2026-04-10

### [What Chinese Dialects Does VoxCPM2 Support? A Complete Regional Speech Guide](/OpenBMB/VoxCPM/what-chinese-dialects-does-voxcpm2-support)

Discover the 9 Chinese dialects VoxCPM2 supports: Sichuanese, Cantonese, Wu & more. Learn about its unique tokenizer-free architecture for regional speech processing.

- Tags: api-reference
- Published: 2026-04-10

### [How VoxCPM Handles 30 Languages Without Language Tags: A Technical Deep Dive](/OpenBMB/VoxCPM/how-does-voxcpm-handle-30-languages-without-language-tags)

Discover how VoxCPM processes 30 languages without language tags. Learn technical details on implicit language inference using MiniCPM-4 and its unique approach to multilingual text.

- Tags: deep-dive
- Published: 2026-04-10

### [How Much Data Is Needed for Effective VoxCPM Fine-Tuning?](/OpenBMB/VoxCPM/how-much-data-is-needed-for-effective-voxcpm-fine-tuning)

Discover how little data is needed for effective VoxCPM fine-tuning. Achieve high-quality speaker adaptation with as little as 5-10 minutes of clean audio.

- Tags: performance
- Published: 2026-04-10

### [How to Fine-Tune VoxCPM for Custom Speaker Adaptation Using LoRA: A Complete Guide](/OpenBMB/VoxCPM/how-to-fine-tune-voxcpm-for-custom-speaker-adaptation-using-lora)

Master VoxCPM custom speaker adaptation with LoRA. This guide shows fast, memory-efficient fine-tuning by training only low-rank matrices to adapt your voice models.

- Tags: how-to-guide
- Published: 2026-04-10

### [Differences Between VoxCPM2, VoxCPM1.5, and VoxCPM-0.5B: Architecture and Capabilities](/OpenBMB/VoxCPM/differences-between-voxcpm2-voxcpm1.5-and-voxcpm-0.5b-model-versions)

Explore VoxCPM2, VoxCPM1.5, and VoxCPM-0.5B differences. Discover their architecture and capabilities from OpenBMB for advanced multilingual TTS, voice cloning, and edge applications.

- Tags: deep-dive
- Published: 2026-04-10

### [Nano-vLLM: Achieving RTF ~0.13 for Production Speech Synthesis in VoxCPM](/OpenBMB/VoxCPM/what-is-nano-vellm-and-its-rtf-0.13-for-production)

Discover Nano-vLLM, an inference engine for VoxCPM, achieving 0.13 RTF for real-time speech synthesis. Experience 2-3x lower latency for production workloads on RTX 4090.

- Tags: performance
- Published: 2026-04-10

### [How to Optimize VoxCPM Inference Speed for RTF ~0.3 on Consumer GPUs](/OpenBMB/VoxCPM/how-to-optimize-voxcpm-inference-speed-for-rtf-0.3-on-consumer-gpus)

Optimize VoxCPM inference speed to RTF ~0.3 on consumer GPUs using torch compile, half-precision, fewer diffusion steps, and inference mode. Get faster AI generation now.

- Tags: performance
- Published: 2026-04-10

### [Voice Attributes Controlled by Text Prompts in VoxCPM: Complete Guide](/OpenBMB/VoxCPM/what-voice-attributes-can-be-controlled-by-text-prompts-in-voxcpm)

Discover how VoxCPM lets you control voice gender age tone emotion pace and style with simple text prompts eliminating the need for reference audio. Master voice generation today.

- Tags: how-to-guide
- Published: 2026-04-10

### [How to Create Custom Voices from Text Descriptions in VoxCPM Voice Design](/OpenBMB/VoxCPM/how-to-create-custom-voices-from-text-descriptions-in-voxcpm-voice-design)

Learn to create custom voices from text descriptions with VoxCPM Voice Design. Generate novel voices using control instructions and the VoxCPM2 model without reference audio.

- Tags: how-to-guide
- Published: 2026-04-10

### [How to Maximize Voice Cloning Similarity with Reference and Prompt Audio in VoxCPM](/OpenBMB/VoxCPM/how-to-maximize-voice-cloning-similarity-with-reference-and-prompt-audio)

Maximize voice cloning similarity in VoxCPM. Learn to use reference and prompt audio, plus essential preprocessing, for superior voice replication.

- Tags: how-to-guide
- Published: 2026-04-10

### [How `prompt_wav_path` and `reference_wav_path` Affect Voice Cloning Quality in VoxCPM](/OpenBMB/VoxCPM/how-do-prompt-wav-path-and-reference-wav-path-affect-voice-cloning-quality)

Discover how prompt_wav_path and reference_wav_path influence VoxCPM voice cloning quality. Learn how to optimize audio inputs for superior synthesis and style control.

- Tags: deep-dive
- Published: 2026-04-10

### [Controllable Voice Cloning vs Ultimate Cloning in VoxCPM2: Key Differences and Implementation](/OpenBMB/VoxCPM/difference-between-controllable-voice-cloning-and-ultimate-cloning-in-voxcpm2)

Explore Controllable Voice Cloning vs Ultimate Cloning in VoxCPM2. Understand how each method preserves timbre or reproduces every vocal nuance for your AI voice projects.

- Tags: deep-dive
- Published: 2026-04-10

### [How VoxCPM Leverages the MiniCPM-4 Backbone for Speech Generation](/OpenBMB/VoxCPM/what-is-the-relationship-between-voxcpm-and-the-minicpm-4-backbone)

Discover how VoxCPM system uses the MiniCPM-4 backbone for advanced speech generation. Explore its powerful text-to-speech capabilities.

- Tags: deep-dive
- Published: 2026-04-10

### [How AudioVAE V2 Achieves Native 48kHz Speech Output in OpenBMB/VoxCPM](/OpenBMB/VoxCPM/how-does-audiovael-v2-achieve-native-48khz-speech-output)

Discover how AudioVAE V2 generates native 48kHz speech with OpenBMB/VoxCPM. Learn about learned transformations and ditching external resampling for pure audio.

- Tags: deep-dive
- Published: 2026-04-10

### [Understanding the VoxCPM Pipeline: LocEnc → TTSLM → RALM → LocDiT](/OpenBMB/VoxCPM/what-is-the-voxcpm-pipeline-locenc-tts-lm-ralm-locdit)

Explore the VoxCPM pipeline: LocEnc, TTSLM, RALM and LocDiT. Understand how these modules process speech and audio for advanced text-to-speech generation and refinement via OpenBMB.

- Tags: deep-dive
- Published: 2026-04-10

### [How VoxCPM's Tokenizer-Free Diffusion Autoregressive Architecture Works](/OpenBMB/VoxCPM/how-does-voxcpm-tokenizer-free-diffusion-autoregressive-architecture-work)

Discover how VoxCPM's tokenizer free diffusion autoregressive architecture generates speech directly from text to audio using a character level language model and local diffusion transformer.

- Tags: internals
- Published: 2026-04-10