
VoiceStudio
The open-source ElevenLabs alternative for local voice cloning, design, create, dubbing and dictation Desktop App
The Lens
By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.
Updated Nov 2026
VoiceStudio (formerly OmniVoice Studio) is a local answer to ElevenLabs. It clones a voice from a clean reference clip, designs new voices from a text description, dubs video into other languages, narrates audiobooks and runs a dictation widget. The desktop app is AGPL-3.0, free, and needs no account.
A GPU is optional: it runs on the CPU, just slower, from about 8 GB of RAM and 10 GB of disk. NVIDIA and Apple Silicon are the smooth paths; AMD support is partial and partly experimental. It exposes a local API and an MCP server. Intel Macs need a backend on another machine.
VoiceStudio Pro is a separate closed-source build: $99 per user per year at launch ($199 list), or $199 one-time for Lifetime, with a commercial license for client work, three devices and remote GPU workers. Solo: free. Small teams: free, Pro if you sell voice work. Larger teams: ask about Enterprise.
The catch is the models, not the app. The default engine's weights are CC-BY-NC (noncommercial), and the project's own license inventory clears none of its models for commercial use, Pro included. Read each model card before paid work, and clone only voices you have permission to use.
Free vs Self-Hosted vs Paid
open coreFree: The AGPL-3.0 desktop app, with no account and no cloud: voice cloning, voice design, video dubbing, dictation and transcription, audiobook creation, a local API and an MCP server. Commercial use of the app is allowed under AGPL terms. Model weights carry their own licenses: the default OmniVoice weights are CC-BY-NC (noncommercial), while model cards for CosyVoice 3, VoxCPM2 and MOSS-TTS list Apache-2.0. The project's own inventory of 63 model repositories marks none as cleared for commercial use, so those labels are upstream metadata you need to check yourself.
Self-hosted: Runs on your hardware. Desktop builds cover macOS 13.3+ on Apple Silicon, Windows 10/11 x64 (Windows on ARM experimental) and Linux x64, with preview Android 10+ and iOS 17+ apps. Plan on about 8 GB of RAM and roughly 10 GB of free disk; a GPU is optional but much faster. NVIDIA CUDA and Apple Silicon are the main paths; AMD has ROCm on Linux, an experimental Windows ROCm setup for selected RX 6000 cards, and Vulkan for GGUF engines. Docker images cover CPU, CUDA and ROCm on amd64, plus an arm64 NVIDIA CUDA build. Intel Macs run only the UI and need a backend on another machine.
Paid: VoiceStudio Pro is $99 per user per year at launch (list $199): 1 user, 3 active devices, a Commercial License for paid client work, remote GPU workers and early access. Lifetime is $199 one-time at launch (list $499) with the same limits. Both are a closed-source build that requires an account and sends license and usage data; 7-day full refund, and Pro does not auto-renew. Enterprise is custom. Speech model licenses stay separate from Pro. VoiceStudio Cloud, a hosted voice and dubbing API, is not available yet; early access offers free usage credits.
The AGPL app is free with every core feature. Pro ($99/user/yr, or $199 lifetime at launch) adds a commercial license, multi-device use and remote GPUs, but neither clears the default voice model's noncommercial weights.
What to do by team size
- Solo
- free; check the model license before paid voice work
- Small team
- free, or Pro at $99/user/yr if you want a commercial license for client work
- Medium team
- free for internal use; Pro per user for client work
- Large team
- Enterprise by enquiry, plus your own model license review
Get tools like this every Wednesday
One featured tool, three on the radar. No fluff.
Similar Tools

Open-source AI engine, run any model locally

the subtitle editor :)

Buzz transcribes and translates audio offline on your personal computer. Powered by OpenAI's Whisper.

The open-source screen recorder and editor for professional product videos, demos, and tutorials.

macOS video editor built for AI

Effort free video editing!
A low score is not a verdict on quality. Young and niche tools start low by design. How we calculate scores
Trust Signals
License: GNU Affero General Public License v3.0
App code is AGPL-3.0 and allows commercial use; model weights carry separate licenses, and the default model is noncommercial.
Commercial use: ✓ Yes
About
- Owner
- Palash Debnath (User)
- Stars
- 111,577
- Forks
- 12,374
Explore Further
More tools in the directory
ollama
Get up and running with Kimi-K2.5, GLM-5, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
187.2k ★Stirling-PDF
#1 PDF Application on GitHub that lets you edit PDFs on any device anywhere
96.7k ★unsloth
Unsloth is a local UI for training and running Kimi K3, Gemma 4, Qwen3.6, DeepSeek-V4, GLM and other models.
79.6k ★Featured in The Open Source Drop #24