Google Cloud Text-to-Speech vs VoxCPM

Google Cloud Text-to-Speech is an API that converts text or SSML into audio data for applications. It offers more than 380 voices across more than 75 languages and variants, with REST or gRPC integration and output formats including MP3, LINEAR16, OGG_OPUS, MULAW, and ALAW. Gemini-TTS supports single- or multispeaker speech guided by natural-language prompts for style, accent, pace, tone, and emotional expression. Chirp 3 instant custom voice can create a personalized voice model from as little as 10 seconds of audio and is available in more than 30 locales. Streaming synthesis supports real-time conversations, while long-audio synthesis accepts up to 1 million bytes of input. Speech controls include pitch, speaking rate, and volume gain. Billing counts input characters, including spaces, newlines, and SSML tags except the mark tag, and billing must be enabled. Standard and WaveNet voices cost US$4 per 1 million characters, with 0 to 4 million characters free. Other voice models have different rates and free-use limits. New customers can receive $300 in Google Cloud credits.

VoxCPM is an open-source text-to-speech system whose VoxCPM2 model generates expressive multilingual speech. It supports 30 languages and lists nine Chinese dialects. Voice Design can create a voice from a natural-language description without reference audio, and voice cloning is supported. Output is 48kHz audio, reference audio is accepted at 16kHz, and WAV is listed as an export format. A streaming generation API is available; the repository reports a real-time factor as low as about 0.3 on an NVIDIA RTX 4090. Full and LoRA fine-tuning can adapt the model from 5–10 minutes of audio. Documented deployment includes Python and CLI use, a local web demo, production serving and on-device inference. vLLM-Omni exposes an OpenAI-compatible /v1/audio/speech endpoint, while Nano-vLLM supports concurrent requests. Apple Silicon MPS and C++ inference on CPU, Metal, CUDA or Vulkan are documented. Code and weights use Apache-2.0 and are free for commercial use. Python installation requires Python 3.10 or later but below 3.13, PyTorch 2.5.0 or later and CUDA 12.0 or later. Voice Design and cloning results can vary, and unsupported languages may require testing or fine-tuning. VoxCPM is headquartered in Beijing, China.