config.yaml

readme: |
  # Fish Diffusion - HiFiSinger Demo 🎤
  GitHub Repo: [fishaudio/fish-diffusion](https://github.com/fishaudio/fish-diffusion) 
  To share a new model, please check out the [Share Your Model](https://huggingface.co./spaces/fishaudio/fish-diffusion/discussions/2) discussion.
max_mixing_speakers: 3

models:
  - name: "M4Singer Pretrained (Many Speakers, Alto, Tenor, Soprano, Bass)"
    config: configs/M4Singer.py
    checkpoint: checkpoints/M4Singer.ckpt
    readme: |
      This model is trained on the Opencpop and M4Singer dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It contains more than 20 speakers and is thus a good playground for timbre mixing.
    default_speaker: "opencpop"
    
  - name: "Tohoku Kiritan (Feminine)"
    config: configs/Kiritan.py
    checkpoint: checkpoints/Kiritan.ckpt
    readme: |
      This model is trained on the Tohoku Kiritan dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a cute, yet powerful voice. 
    default_speaker: "kiritan"

  - name: "Tohoku Itako (Feminine)"
    config: configs/Itako.py
    checkpoint: checkpoints/Itako.ckpt
    readme: |
      This model is trained on the Tohoku Itako dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a bright and whispery voice. 
    default_speaker: "itako"

  - name: "No.7 (Feminine)"
    config: configs/Seven.py
    checkpoint: checkpoints/Seven.ckpt
    readme: |
      This model is trained on the No.7 dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a strong and sharp voice.
    default_speaker: "seven"

  - name: "Yoko (Feminine)"
    config: configs/Yoko.py
    checkpoint: checkpoints/Yoko.ckpt
    readme: |
      This model is trained on the Sinsy-f00001 dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a powerful, tense, and relaxed voice.
    default_speaker: "yoko"

  - name: "JSUT (Feminine)"
    config: configs/JSUT.py
    checkpoint: checkpoints/JSUT.ckpt
    readme: |
      This model is trained on the JSUT-song dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a moist and transparent voice.
    default_speaker: "jsut"

  - name: "CSD (Feminine)"
    config: configs/CSD.py
    checkpoint: checkpoints/CSD.ckpt
    readme: |
      This model is trained on the Children's Song Dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a sweet and tender voice.
    default_speaker: "csd"

  - name: "Namine Ritsu (Feminine)"
    config: configs/Ritsu.py
    checkpoint: checkpoints/Ritsu.ckpt
    readme: |
      This model is trained on the Namine Ritsu ENUNU Dataset and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a powerful and throaty voice.
    default_speaker: "ritsu"

  - name: "S (Masculine)"
    config: configs/S.py
    checkpoint: checkpoints/S.ckpt
    readme: |
      This model is trained on a datset known as S and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a nasally but powerful voice.
    default_speaker: "s"

  - name: "C (Feminine)"
    config: configs/C.py
    checkpoint: checkpoints/C.ckpt
    readme: |
      This model is trained on a datset known as C and released under the [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
      It has a whispery, fluttery voice.
    default_speaker: "c"