VoxCPM2, in Studio Quality
VoxCPM2 clones a voice or designs one from a plain-language description, outputs studio-quality 48kHz audio, and speaks 30 languages. Try it in the browser or offline on desktop.
VoxCPM2 in Voice Creator Pro
Hear a voice designed from text
VoxCPM2 builds a voice from a plain-language description, with no reference audio. Every voice below was generated from the exact description shown next to it.
Tap to hear a voice designed from its description
No reference audio
Documentary Narrator
“A middle-aged man with a calm, resonant, low-pitched voice, smooth with a slight gravelly warmth. Speaks slowly and deliberately with measured pauses, perfect for a nature documentary.”
Movie Trailer
“A man with a deep, booming, heavily resonant voice and a rough gravelly edge. Delivers slowly at high volume with long dramatic pauses between phrases, perfect for an epic movie trailer.”
Late-Night Radio Host
“A woman in her thirties with a low-pitched, velvety voice, breathy and warm with a relaxed rasp. Speaks slowly and intimately at low volume, suitable for late-night radio.”
Panicked Boy
“A young boy with a thin, high-pitched voice, light and slightly strained. Speaks fast and urgently at rising volume, panicked and scared, for a horror game.”
Written like casting notes
The best descriptions stack three layers. Here is how one reads, from who is speaking to how it performs.
An elderly woman.
Quiet and raspy, low-pitched, with a distinct grainy texture and subtle breathy tremors.
A slow tone at a very low volume, perfect for historical narration.
The full description
A quiet raspy, elderly woman of a low-pitched voice with a distinct, grainy texture and subtle breathy tremors. Delivers a slow tone at a very low volume, perfect for historical narration.
Why VoxCPM2
Why people reach for VoxCPM2
VoxCPM2 is the model to reach for when audio quality and language coverage matter. It outputs studio-quality 48kHz audio directly, speaks 30 languages from a single model, and both clones a voice from a reference and designs one from a detailed description. Its clones score among the strongest for speaker similarity on the standard benchmarks.
Voice Creator Pro runs it for you, in the browser or offline on the desktop, and ships starter presets, from broadcaster and documentary narrator to extreme emotions, so you can start from a working description and edit rather than write from scratch.
Studio-Quality 48kHz
Outputs 48kHz audio directly, with no external upsampler, where most models output at a lower sample rate.
30 Languages
Speaks 30 languages from a single model, plus nine Chinese dialects, so a designed voice is not stuck in English.
Scene-Defined Voices
Name the setting, like a trailer or late-night radio, and VoxCPM2 shapes the whole performance to fit it.
High-Similarity Cloning
Clone a voice from a reference clip. VoxCPM2 scores among the top open models for speaker similarity on the standard benchmarks.
Cloud or Desktop
Two ways to use VoxCPM2
Both give you the full VoxCPM2 model with layered voice design, cloning, and 48kHz output. Pick the one that fits how you work.
Voice Creator Pro Cloud
Run it in your browser
- Nothing to install, works from any laptop or phone
- No GPU needed, server GPUs run the model for you
- Free tier that renews monthly, and paid plans that scale with your needs
- Best if you want to start instantly with nothing to install
Desktop App
Run it locally and offline
- Runs VoxCPM2 on your own machine, 100% offline
- Unlimited generations, one-time purchase from $54.99
- Your text and audio never leave your device
- Runs on a mainstream GPU, VoxCPM2 needs around 6 GB VRAM
Get Started
How it works
The free browser tier is the quickest start. Prefer to work offline? The same steps apply in the desktop app.
Create a free account
Sign up for Voice Creator Pro Cloud in your browser, or install the desktop app. The free tier gives you a monthly allowance with no card required.
Describe or preset your voice
Write a layered description, identity, texture, and delivery, or start from a preset and edit it. When a designed voice lands, save its seed to reuse it.
Type your text and generate
Enter your script, select VoxCPM2 as the model, and generate 48kHz audio ready to download.
Use Cases
Who it's for
Narrators and Documentary
Produce narration where audio fidelity is front and center. The 48kHz output holds up in a mix without sounding thin.
Multilingual Creators
Design one voice and use it across 30 languages, so a series or brand voice is not confined to English.
Trailers and Scene Work
Define a voice by its scene, a trailer, a bedtime story, late-night radio, and let the description shape the delivery.
Studios and Power Users
Run VoxCPM2 offline in the desktop app for unlimited, private generation on your own hardware.
Best results
Tips for best results
Stack the three layers
Cover identity, texture, and delivery in your description. The more of the voice you specify, the closer VoxCPM2 lands, and naming the setting steers the whole performance.
Match the text to the emotion
VoxCPM2 also reads the words you give it. A furious description paired with a calm sentence underdelivers, so give an intense voice something worth the intensity.
Lock the seed once it lands
Designs vary between takes, so plan on a few attempts. When one nails it, reuse its seed and save the voice. Our VoxCPM2 voice design guide has the full three-layer format and more example descriptions.
Voice Creator Pro
Get started with VoxCPM2
Start free in your browser today, or buy the desktop app once and run VoxCPM2 offline with unlimited generations. Both include layered voice design, cloning, 48kHz output, and full commercial rights.
Studio-Quality 48kHz
Outputs 48kHz audio directly with no upsampler, for narration and other quality-first work
30 Languages
Speaks 30 languages from a single model, plus nine Chinese dialects, so a voice is not limited to English
Layered Voice Design
Describe a voice in identity, texture, and delivery layers, the way you would write casting notes
High-Similarity Cloning
Clone a voice from a reference clip, with top-tier speaker similarity on the standard benchmarks
Cloud or Offline Desktop
Run it free in your browser, or offline on the desktop app with unlimited generations
Commercial License
You own every clip you generate, with full commercial rights and no attribution
FAQ
Common Questions
VoxCPM2 is a voice model in Voice Creator Pro built for studio-quality voice design. You describe a voice in plain language and it builds one, it outputs 48kHz audio directly, and it speaks 30 languages from a single model. It also supports voice cloning and streaming.
Yes. The Voice Creator Pro Cloud free tier lets you design a voice and hear VoxCPM2's 48kHz output in your browser, with no card required. Cloud plans start at $5 a month, and the desktop app runs VoxCPM2 offline with unlimited generations as a one-time purchase from $54.99.
It outputs studio-quality 48kHz audio directly, with no external upsampler, while most models output at a lower sample rate. That higher fidelity is why it suits narration, documentary, and other work where audio quality is front and center.
VoxCPM2 speaks 30 languages from a single model, so a designed voice is not limited to English. It also handles nine Chinese dialects.
You describe it in plain language, and the best results come from stacking three layers: identity (who is speaking), texture (what the voice is made of, like low-pitched or raspy), and delivery (how it performs, and where it belongs). Naming the setting, like perfect for a movie trailer, steers the whole performance. Voice Creator Pro also ships starter voice design presets you can edit.
Designed voices vary between generations, so plan on two or three attempts. When one nails it, reuse its seed to reproduce that exact take, then save it as a voice. From there you can clone it across any text with the timbre locked in.
Yes. Alongside voice design, VoxCPM2 does zero-shot voice cloning from a reference clip, and it scores among the top open models for speaker similarity on the standard cloning benchmarks. A common workflow is to design a voice first, lock its seed, and then clone that saved voice across any text.
Not for the cloud. In the desktop app, VoxCPM2 runs on your own machine and needs around 6 GB of VRAM, which most recent graphics cards have. If you would rather not install anything, or your machine is lighter, Voice Creator Pro Cloud runs the model on server GPUs so any browser works.
Yes, with the Voice Creator Pro desktop app. It runs VoxCPM2 on your own machine with no connection, unlimited generations, and your 48kHz audio stays local. Voice Creator Pro Cloud is server-based and needs a connection.
Yes. The 48kHz audio you generate with VoxCPM2 is yours, on both Cloud and desktop and including the free tier, with full commercial rights, no royalties, and no attribution required.