VOICEVOX
: How to use it, features, and the business problems it solves

What is VOICEVOX?

VOICEVOX is a free text-to-speech and singing-voice synthesis application developed by Hiho (Kazuyuki Hiroshiba). Born in Japan as an open-source project, it turns typed text into speech using 46 characters (as of August 19, 2026) such as Zundamon and Shikoku Metan, and can be used for both commercial and non-commercial purposes free of charge, provided you include a credit notice.

It is widely used for narration in explainer videos, game commentary, and VTuber content — in Japan, "Zundamon's voice" has practically become synonymous with VOICEVOX. The editor lets you fine-tune intonation character by character, includes "Humming" (talk-like singing) and "Song" (full singing) modes, and runs locally on Windows, Mac, and Linux.

How to use

1. Download from the official site — Get the installer from the official site. A CPU edition and a GPU edition (for supported GPUs) are available. If a security warning appears on first launch, confirm the publisher "Kazuyuki Hiroshiba" and proceed.
2. Enter text — Click the blank area next to a character icon and type the sentence you want read aloud. Each line can have its own text and character.
3. Choose a character — Switch speakers from the icon on the left of the text field. Many characters offer multiple styles such as "Normal" or "Sweet."
4. Play and adjust — Preview with the play button, then adjust per-character pitch bars in the "Intonation" panel. Speed, volume, and inflection can be saved as presets.
5. Export — Use "File" → "Export audio" to save WAV files. Text files and lab (timing) files can be exported together via settings.

Key features

• Text-to-speech — Reads Japanese text aloud with 46 character voices, many with multiple emotional styles
• Intonation editing — Adjust pitch per character with bar controls; devoicing of sounds like "ki," "tsu," and "su" can be toggled
• Song and Humming — Singing-voice synthesis via a piano roll with notes and lyrics; "Humming" sounds talk-like while "Song" sounds sung
• Pronunciation & accent dictionary — Register readings and accents for proper nouns and technical terms
• Presets — Save parameter sets (speed, inflection, etc.) and apply them per line
• Offline operation — Synthesis runs entirely on your PC; no internet connection required
• Engine and core for developers — The synthesis engine is exposed as an HTTP API (AudioQuery) and as an embeddable core library; a multi-engine feature can load third-party engines compatible with the VOICEVOX API

Pricing

VOICEVOX has no paid plan; every feature and every character is free (official statement as of August 19, 2026).

Free plan / trial

There is no free-vs-paid distinction — everything is available the moment you download it, and WAV export is unlimited. The "cost" is contractual rather than monetary: you must include a credit notice showing that VOICEVOX was used (in a video description or within the video) and follow each character's terms of use.

Watch out: terms differ per character

This is the most commonly misunderstood part of VOICEVOX. As the official Q&A explains, the characters and the software have different rights holders, and published audio must follow each character's voice-library terms.

• Credit format — For Tohoku Zunko Project characters such as Zundamon and Shikoku Metan, credits in the form "VOICEVOX:Zundamon" are required. The official Q&A suggests placing them in the video description or the video itself, or at the start/end of audio-only content
• Commercial use — Under that project's voice-library terms, commercial and non-commercial use are both permitted with credit; commercial use without credit requires a contract of JPY 400,000 (excl. tax) per character
• Examples of prohibited uses — Content against public order and morals, political or religious activity, information-product (get-rich-quick) sales, false or misleading content, use by antisocial forces, etc. (examples from the Tohoku Zunko Project terms)
• Every character is different — The above is only one example; the 46 characters have different rights holders and terms. Always check each character's terms on the official site before publishing. When morphing multiple voices, the stricter terms take precedence

Japanese language support

• UI — Fully Japanese. The software, documentation, and official Q&A are all in Japanese
• Input/output — Specialized in reading Japanese text aloud, with features focused on Japanese speech quality such as accent-phrase editing and devoicing control
• Support — Official Q&A page, plus the "#VOICEVOX" hashtag and official account on X. There is no dedicated enterprise support desk

Reputation and adoption

• GitHub stars — The editor repository (VOICEVOX/voicevox) has roughly 3,200 stars (as of August 19, 2026), with active development across the editor, engine, and core repositories
• Sustained OSS development — Updated continuously since its release in August 2021; the latest version as of August 2026 is 0.25.2, with community contributors
• Cultural reach — Videos featuring VOICEVOX characters, especially Zundamon, are firmly established on YouTube and Niconico (no official usage statistics are published)

How it compares

VOICEVOX is optimized for one thing: making Japanese content with character voices, for free. If you need multilingual narration or natural English speech, compare ElevenLabs; if you want business-grade licensing handled in one place, CoeFont; if you want to change your own voice live, Voice.ai. Conversely, if your goal is videos built on the popularity of characters like Zundamon, there is practically no alternative to VOICEVOX.

FAQ

Q. Can I use it commercially? Yes. The official Q&A states that audio created with VOICEVOX can be used commercially or non-commercially as long as you include a credit notice. Each character also has separate voice-library terms, so always check the terms for the character you use.

Q. How should I write the credit? Use a format that identifies both VOICEVOX and the character, such as "VOICEVOX:Zundamon," in the video description or within the video. For audio-only content, the official Q&A suggests inserting a spoken credit at the beginning or end.

Q. Can I use it in monetized YouTube videos? Yes, with a credit notice and within the character's terms (monetization counts as commercial use). Mind each character's prohibited uses, such as political/religious content or content against public order and morals.

Q. Does it work offline? Yes. Synthesis runs locally on your PC with no internet connection required.

Q. Can I embed it in my company's product or service? Technically yes — the engine and core are public. But each character's voice terms differ, and some require a paid contract for use without credit, so confirm with the rights holders.

Q. Does it run on Mac? Yes, on macOS 14 or later (Rosetta may be required on Apple Silicon). Linux (Ubuntu 22.04/24.04) is also supported.

Provider

VOICEVOX is not a company product: it is developed by Hiho (Kazuyuki Hiroshiba) as an individual and maintained together with OSS community contributors. It is published on GitHub in three layers — editor (Electron + Vue), engine (Python), and core (Rust) — with the editor dual-licensed under LGPL-3.0. The project is funded by optional donations via pixivFANBOX. Character rights belong to each character's rights holder (Zundamon and others belong to the Tohoku Zunko Project) and are managed separately from the software.

---

The information on this page is based on what is published on the official website as of August 2026. Features and pricing may change, so please check the official website for the latest information.