jf_alpha · female · 1.0× speed · 2.5 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
Japanese voice comparisons
Compare voices on the same sentence, then compare the pace of one voice. This is a listening comparison only. No voice has been selected or approved. Tap play whenever you’re ready.
Voice comparison
The same route sentence at normal speed. Compare the exact voice names below.
jm_kumo · male · 1.0× speed · 2.6 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
jf_gongitsune · female · 1.0× speed · 3.1 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
jf_nezumi · female · 1.0× speed · 3.0 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
jf_tebukuro · female · 1.0× speed · 2.8 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
Speed comparison
The same female voice, jf_alpha, at 1.0× and 0.85× speed. These are two speeds of one voice.
jf_alpha · female · 1.0× speed · 2.5 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
jf_alpha · female · 0.85× speed · 2.8 seconds
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
Additional sentence examples
Casual past and an invitation, both with female jf_alpha at normal speed.
jf_alpha · female · 1.0× speed · 2.3 seconds
昨日、友だちと映画を見に行った。
Yesterday I went to see a movie with a friend.
jf_alpha · female · 1.0× speed · 4.3 seconds
一緒に昼ごはんを食べませんか。いいですね。じゃあ、あの店に行きましょう。
Shall we have lunch together? Sounds good. Then let's go to that restaurant.
Voice approval is still pending. Lesson audio will wait for your choice.
Generated locally with Kokoro-82M v1.0 via kokoro-onnx 0.6.1. Official voice metadata.
Style-Bert-VITS2 JP-Extra audition
A separate engine, with two JP-Extra voices reading the station sentence from the Kokoro comparison. The lunch clip uses three separate calls: F1 invites, M1 responds, then F1 finishes. The turns are joined by explicit 0.45-second silences, with no speed change. This is a listening audition only. No voice has been selected or approved.
Neutral style at length 1.0, natural speed.
Same station sentence
Two JP-Extra voices read the exact route sentence used by the Kokoro comparison. This is a separate engine, so listen for a different character rather than a ranking.
jvnv-F1-jp · female · Neutral · length 1.0 (natural speed) · 3.0 seconds · 44.1 kHz
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
jvnv-M1-jp · male · Neutral · length 1.0 (natural speed) · 3.1 seconds · 44.1 kHz
すみません、駅までどう行けばいいですか。
Excuse me, how should I go to the station?
Two-speaker lunch exchange
Three calls: F1 invites, M1 responds, then F1 finishes. The turns are joined by explicit 0.45-second silences with no speed change.
jvnv-F1-jp + jvnv-M1-jp · female + male · Neutral · length 1.0 (natural speed) · 5.7 seconds · 44.1 kHz
一緒に昼ごはんを食べませんか。いいですね。じゃあ、あの店に行きましょう。
Shall we have lunch together? Sounds good. Then let's go to that restaurant.
Turn boundaries: F1 invitation, 0.45 s pause, M1 response, 0.45 s pause, F1 final.
Sources, licenses and changes
Generated locally with Style-Bert-VITS2 2.7.0 (source code AGPL-3.0), using the official jvnv-F1-jp and jvnv-M1-jp JP-Extra models (model weights CC BY-SA 4.0) trained on the JVNV corpus (corpus CC BY-SA 4.0). The JVNV corpus was created by Detai Xin, Junfeng Jiang, Shinnosuke Takamichi, Yuki Saito, Akiko Aizawa and Hiroshi Saruwatari. The corpus license is CC BY-SA 4.0.
The AGPL-3.0 code license and the CC BY-SA 4.0 model and corpus licenses apply separately. The developer requests in the official Style-Bert-VITS2 terms that precede the model section are non-binding requests, not new enforceable license terms.
These original audition recordings are offered under CC BY-SA 4.0 to the extent Reibun holds rights in them. Credit Reibun and the model and corpus sources above. Changes: three synthesized turns were concatenated with 0.45-second silences and a uniform −1 dB gain; no speed, pitch or intonation change was made. Share adaptations alike and keep this attribution. This offer does not relicense Reibun, the model weights, the corpus or other recordings.
Voice approval is still pending. These files are an audition, not a chosen voice.