Ready-made app · VoxCPM2

Run VoxCPM2 on a rented GPU

OpenBMB’s VoxCPM2 turns text into speech in 30 languages with lively intonation. Describe a voice in words or take it from a short recording. It runs on a 12 GB card, and the interface opens in your browser.

App: Text to speech · VoxCPM2
GPU: RTX 3060, 12 GB
Price: from €0.12/h

Launch VoxCPM2 · from €0.12/h Other servers

To launch, your balance must cover one hour of the machine. You are billed by the minute for actual use. The server installs Text to speech · VoxCPM2 for you.

A studio microphone whose sound wave turns into neat rows of transcript lines

from 12 GBof video memory

2–5 minutesuntil the server is up

By the minutebilled from your balance

Gradioopens in your browser

Examples: what went in and what came out

The same texts, voiced in the ready-made app without any processing. Press play to listen.

EN

Welcome to the product tour. In the next two minutes you'll learn how to launch your first GPU server, connect to it, and run a model. No setup required. Let's get started.
EN

Once upon a time, in a small village by the sea, there lived an old lighthouse keeper who knew every star by name. Each night he climbed the stairs and told the sky a new story.

Launch VoxCPM2 · from €0.12/h RTX 3060, 12 GB · billed by the minute

What is on the server after launch

Text to speech · VoxCPM2 is free: you only pay for GPU time. Everything installs at launch; there is nothing to configure by hand.

Web interface in your browser
VoxCPM2, Apache 2.0 licence
A button in the dashboard that opens Gradio
Your folder with models and results can be saved when you stop

Licence Apache 2.0: you can use the speech commercially.

How much memory you need and which GPU to pick

Text to speech · VoxCPM2 needs a card with at least 12 GB; the cheapest suitable one is the RTX 3060. The table shows memory for other versions and tasks. Prices update from available catalogue machines when you open the page.

Version	Video memoryMemory	Suitable GPUGPU	Price per hourPer hour
VoxCPM2, ready-made app	from 12 GB	RTX 3060	€0.12

Guide from real launches: complex graphs and high resolutions need more memory. Pick: the cheapest single card with enough memory. All GPUs and prices

How to run VoxCPM2 in 3 steps

Sign up and top up. Create an account and top up from €10. To launch, your balance must cover at least one hour of the chosen machine; after that you pay by the minute.
Launch the app. Click Launch VoxCPM2. The dashboard opens with Text to speech · VoxCPM2 and a card with at least 12 GB already selected; check the price and confirm.
Work in your browser. The server starts in 2–5 minutes, and the app downloads the model in a few more. Then press the Gradio button in the dashboard and type your prompt.

Finished? Stop the server in the dashboard and billing stops. Save and stop keeps /workspace for €0.03 per GB per month, and it comes back on your next server automatically.

Commands for a manual install on an empty server

Install on an empty server

git clone --depth 1 --branch 2.0.3 https://github.com/OpenBMB/VoxCPM && cd VoxCPM pip install voxcpm==2.0.3 torchcodec==0.5 python app.py --port 7860

What people run VoxCPM2 for

Audiobooks and podcasts

Long texts with natural pauses and intonation.

Speech in many languages

One video in versions for different countries.

A voice from a description

For example, "a calm, low male voice", without any recording of your own.

Tips before you launch

For your own voice, upload a clean 5–10 second recording together with its transcript.
Punctuation shapes the intonation: the model reads exclamations and questions differently.

Frequently asked questions

Which GPU do I need for VoxCPM2?

A 12 GB card such as the RTX 3060 is enough. The model is about 5 GB and downloads during installation.

How is VoxCPM2 different from Qwen3-TTS?

VoxCPM2 sounds more natural and can copy a voice from a sample. Qwen3-TTS is simpler: pick a ready voice from the list and go.

How long does a launch take?

The server usually starts in 2–5 minutes. Then the app downloads the model on its own, which takes a few more minutes, and a button that opens Gradio appears in the dashboard.

What happens to my models and results after I stop?

When you stop, click Save and stop: the /workspace folder with your models, LoRAs and results goes to storage for €0.03 per GB per month and comes back on your next server automatically. If you stop without saving, the disk is deleted.

How is it billed?

By the minute from your balance in euros. Top up from €10; to launch, your balance must cover at least one hour of the chosen machine. Stop the server and billing stops.