Akjava/matcha-tts_ja_100speakers_group006

This model is a 100-speaker multispeaker model in the Matcha-TTS format/architecture.(trained with Japanese)

家具商人のフィシェルは、荷車と仔馬を貸してくれた。

spk10:A lower-pitched female voice with a strong core

私はあなたのことが心配です

spk99:A slightly quirky female voice that leaves a strong impression

僕はいつか面白いゲームを作りたい

spk26:AI-Game-Bu:SEAN

This model is replaced 10 qwen-character to chatterbox(common voice) character.

trained mel_mean/mel_std is difference than group005qw

Qwen3-TTS and Chatterbox Multingual Mixed

Chatterbox I faild to confirm watermark because of technical probrom,but maybe chatterbox watermark is exist. If you don't like the watermark, use qwen3-tts only version

• there are similar qwen3-tts only version

license

This model license is under MIT

My training data is created by Apache Licensed/mit model output. https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base

Matcha-TTS is MIT https://github.com/shivammehta25/Matcha-TTS

Training

need checkpoint and audio from there https://huggingface.co/Akjava/matcha-tts_ja_100speakers_group005qw-100

Use this. https://github.com/akjava/Matcha-TTS-Japanese

Demo

https://ai-game-bu.itch.io/ai-gaming-voice