| š» | Technology article work in progress. For information on how to help, see the guidelines. More subjects categorized here. |
š» |
VOICEVOX is an open source, deep learning, reading/text-to-speech synthesizer software developed by Hiho. It later received singing synthesizer features on January 31st, 2024.
Etymology
VOICEVOX was named out of a desire to make a software that could fit many voices like in a box. The "VOX" part of the name comes from the latin word for "voice", but is also a homonym for "box".[1]
VOICEVOX's main type of synthesis, text-to-speech was rebranded into VOICEVOX Talk to differentiate itself from the new type of synthesis introduced in January 2024, VOICEVOX Song and Humming. According to Hiho, Song was chosen over Sing because, brand-wise a noun sounded better than verb, a problem Talk did not run into since the word is used as both a noun and verb.[2]
Product information
Demonstrations
| Demonstrations |
|---|
|
| |
|---|---|
| Generation 1 demo | YouTube |
| Generation 2 demo | YouTube |
| Generation 3 demo | YouTube |
| Generation 4 demo | YouTube |
Requirements
- OS:
- Windows: Windows 10 / Windows 11
- Mac: macOS Catalina or later
- Linux: Ubuntu 18.04 / Ubuntu 20.04
- GPU: Nvidia
VOICEVOX Talk
VOICEVOX Talk (VOICEVOX ćć¼ćÆ) is VOICEVOX's text-to-speech feature.
Releases
1st generation
2nd generation
3rd generation
4th generation
5th generation
6th generation
7th generation
8th generation
Upcoming voicebanks
(Generation 9)
(Generation 9)
VOICEVOX Song
VOICEVOX Song (VOICEVOX ć½ć³ć°) is VOICEVOX's singing synthesis feature. It was first revealed with the Humming feature on December 28, 2022.[3] A prototype release for Song and Humming was slated for late January 2024. The prototype will only have the bare minimum features available, with pitch editing and other features planned post-release. The user interface is also a prototype, which will be improved based on user feedback. The editor and API are planned to be open source. The editor for Humming will be integrated into the base VOICEVOX software.[4] The prototype was released on January 31, 2024 and properly introduced as VOICEVOX Song.[5]
Humming
Humming (ććć³ć°) is a feature that allows VOICEVOX voices to sing using their TTS voicebanks. Compatibility is planned by generation, with generations 1 and 2 in development at the time of announcement.[6] Theoretically, all voices and emotion styles should be compatible with the Humming feature however, VOICEVOX will have to consult with the IP owners if the output is up to par, if not then those databases will not have the Humming feature available.
Releases
Humming Compatibility
Upcoming voicebanks
VOICEVOX Nemo
VOICEVOX Nemo is VOICEVOX's effort to provide voicebanks without characters and easier licensing. It was first announced on March 18, 2022.[7] The voices were taken from an audition that started on the same day of its announcement and the winners were revealed on April 1, 2022.[8] A demo was showcased on January 2, 2023.[9] but the voices themselves would not end up releasing until November 17,[10] featuring an additional female voice.
Releases
(voiced by äŗå²ęÆ å)
(voiced by éå·ćć)
(voiced by ćć)
(voiced by ć¬ć£ć“ć)
(voiced by ććć ć¾ć)
(voiced by č¤ē°ę代)
(voiced by ć¬ćć¼ćć»ćøć³)
(voiced by ćć”ćć“ćŖć©)
(voiced by å¾ ć”äŗŗ)
References
- ā https://twitter.com/hiho_karuta/status/1421652582921957377
- ā https://twitter.com/hiho_karuta/status/1752674531553292716
- ā https://twitter.com/voicevox_pj/status/1740332179895570484
- ā https://twitter.com/voicevox_pj/status/1740332184022761938
- ā https://twitter.com/voicevox_pj/status/1752652721520521241
- ā https://twitter.com/voicevox_pj/status/1740332184022761938
- ā https://twitter.com/voicevox_pj/status/1504759518873505828
- ā https://twitter.com/hiho_karuta/status/1509840389921935362
- ā https://twitter.com/voicevox_pj/status/1609913075943247879
- ā https://twitter.com/voicevox_pj/status/1725439773475811495
External links
Official
Unofficial
Articles
| ||||||||||||||||||||||||||||||||||