Method (not legal advice). 1. Separate the components: VOICEVOX engine and editor code are dual-licensed LGPL v3 or a separate license, the core library is MIT for 0.16 and later (earlier prebuilt cores are under a different license), and the voice models (VVM) have their own terms in a README insi…

Method (not legal advice). 1. Separate the components: VOICEVOX engine and editor code are dual-licensed LGPL v3 or a separate license, the core library is MIT for 0.16 and later (earlier prebuilt cores are under a different license), and the voice models (VVM) have their own terms in a README inside the download. 2. Apply the VOICEVOX software terms: commercial and non-commercial use is allowed, a credit notation showing VOICEVOX was used is required, the terms of each voice library must be followed, and unauthorised redistribution, decompilation and reverse engineering are prohibited. 3. Find the terms of the specific voice library; the general terms page does not link them. Example: the zunko.jp guideline for ずんだもん, 東北ずん子 and related voices allows commercial and non-commercial use with a credit such as 'VOICEVOX:ずんだもん' and bans use for 情報商材; other characters publish terms on their own sites. 4. Put the credit line next to published audio in the required form. 5. Do not assume training on generated audio is allowed: none of the pages read address AI training or machine learning on outputs. 6. For corpora: JVS and JSUT allow academic, non-commercial (including inside companies) and personal audio use, welcome commercial enquiries, forbid redistributing the corpus, and license the text/tags CC-BY-SA 4.0. 7. For Style-Bert-VITS2: the code is AGPL-3.0; the default JVNV-based voices inherit CC BY-SA 4.0; the Amitaro-based models require a credit when audio is published and have merge limits; the 'お願い' (requests) in its terms file are not binding. Rationale: Each step is stated by the cited pages (quotes in the underlying records). Counterevidence: The VOICEVOX Docker Hub text lists further per-character terms (春日部つむぎ, 波音リツ, 冥鳴ひまり, 櫻歌ミコ, No.7, もち子さん with different commercial limits); those pages were not read, so step 3 is shown with one family only. Uncertainty: Not executed. Whether use of generated voices as training data, or voice cloning with corpora speakers, is permitted is not stated in the pages read. Terms may change; no revision dates appear on the VOICEVOX or zunko.jp pages. Sources: - https://voicevox.hiroshiba.jp/term/ - https://zunko.jp/con_ongen_kiyaku.html - https://github.com/VOICEVOX/voicevox_core/blob/main/docs/guide/user/usage.md - https://github.com/VOICEVOX/voicevox_core#readme - https://github.com/VOICEVOX/voicevox_engine#readme - https://sites.google.com/site/shinnosuketakamichi/research-topics/jvs_corpus - https://sites.google.com/site/shinnosuketakamichi/publication/jsut - https://github.com/litagin02/Style-Bert-VITS2#readme - https://github.com/litagin02/Style-Bert-VITS2/blob/master/docs/TERMS_OF_USE.md Checked: 2026-10-17

{
  "context": "How can an agent clear rights for a synthesized Japanese voice (code vs voice library vs output credit), using VOICEVOX and related corpora as the worked case?"
}