VOICEVOX

Free plan APILinuxMacSelf-hostedWindows

What VOICEVOX is

VOICEVOX is free software for text-to-speech and singing voice synthesis on Windows, macOS, and Linux, with API and self-hosted options. Users can adjust intonation, accents, speaking speed, pitch, volume, and pauses, then export audio as WAV. Multiple character voices are available, with usage rules set for each character. The song feature is a prototype offering humming in a character’s speaking voice and a separate singing mode. VOICEVOX can run on an offline computer. Its engine generates audio for other software and applications and is distributed as executables and a Docker image. VOICEVOX CORE is available as a library for embedding synthesis in applications and services. The project is built on an open-source version and accepts contributions. Generated audio may be used commercially or non-commercially with credit, subject to VOICEVOX and character terms. GPU mode supports DirectML on Windows and Nvidia GPUs on Linux; macOS has no GPU mode.

Who it suits

VOICEVOX suits creators and developers who need controllable speech or singing synthesis, WAV export, offline use, or application integration. It also fits users who can follow credit and character-specific usage conditions.

What it does well

Free for commercial and non-commercial use
Detailed controls for speech output
WAV audio export
Offline computer use supported
Engine, API, Docker, and library integration

What to know first

Character-specific usage rules apply
Song feature remains a prototype
Each note accepts one Japanese character
macOS does not support GPU mode

Plans

Plan	Billed and limits	Price
VOICEVOX	Free for commercial and non-commercial use · character-specific terms apply	Free

As voicevox.hiroshiba.jp lists them, read 12 Oct 2026. The maker's own page has the final word.

Verdict

VOICEVOX offers adjustable speech, prototype singing, offline operation, and several integration routes without a listed price. Users should check each character’s terms and expect limitations in the developing song feature.

Facts from VOICEVOX's own pages

API access: Yes · voicevox.hiroshiba.jp, 20 Sept 2026
Audio export: Generated audio can be exported as WAV files. · voicevox.hiroshiba.jp, 12 Oct 2026
Characters: The site lists multiple character voices, and says usage rules vary by character. · voicevox.hiroshiba.jp, 12 Oct 2026
Commercial use: Yes · voicevox.hiroshiba.jp, 20 Sept 2026
Engine integration: The VOICEVOX engine can generate audio from other software or applications and is distributed as executables and a Docker image. · voicevox.hiroshiba.jp, 12 Oct 2026
Export formats: WAV · voicevox.hiroshiba.jp, 20 Sept 2026
Free plan: Yes · voicevox.hiroshiba.jp, 20 Sept 2026
GPU limits: GPU mode supports DirectML GPUs on Windows and Nvidia GPUs on Linux; macOS does not support GPU mode. · voicevox.hiroshiba.jp, 12 Oct 2026
Library integration: The VOICEVOX CORE library is offered for embedding speech synthesis in applications and services. · voicevox.hiroshiba.jp, 12 Oct 2026
License conditions: The FAQ says generated audio can be used commercially or non-commercially with credit, subject to VOICEVOX and each character's terms. · voicevox.hiroshiba.jp, 12 Oct 2026
Offline use: The FAQ says VOICEVOX can be used on an offline computer. · voicevox.hiroshiba.jp, 12 Oct 2026
Open source: The site says VOICEVOX is built on an open-source version and invites users to contribute improvements. · voicevox.hiroshiba.jp, 12 Oct 2026
Platform requirements: The FAQ lists Windows 10 or 11, macOS 14 or later, and Ubuntu 22.04 or 24.04 for the CPU version. · voicevox.hiroshiba.jp, 12 Oct 2026
Platforms: Windows,macOS,Linux,API · voicevox.hiroshiba.jp, 20 Sept 2026
Purpose: VOICEVOX is software for text-to-speech and singing voice synthesis. · voicevox.hiroshiba.jp, 12 Oct 2026
Singing: The prototype song feature offers humming in a character's speaking voice and a separate singing voice mode. · voicevox.hiroshiba.jp, 12 Oct 2026
Song feature limit: The song feature is described as a prototype under active development; its FAQ says each note currently accepts only one Japanese character and overlapping notes cause errors. · voicevox.hiroshiba.jp, 12 Oct 2026
Speech controls: Users can adjust intonation, accents, speaking speed, pitch, volume, and pauses. · voicevox.hiroshiba.jp, 12 Oct 2026
Support: The FAQ directs users to post questions or requests with #VOICEVOX on X and asks bug reporters to contact the VOICEVOX account there. · voicevox.hiroshiba.jp, 12 Oct 2026

Where VOICEVOX stands on Freedom251

Text-to-speech tools#23 of 232
Text-to-Speech Software#114 of 223