voice_audio_multi/model_specs/kugelaudio.json

{
  "schema_version": 1,
  "family": "kugelaudio",
  "display_name": "KugelAudio-0-Open",
  "description": "KugelAudio text-to-speech with bundled preset voices and incremental audio output.",
  "category": "tts",
  "status": "experimental",
  "tasks": [
    "tts"
  ],
  "modes": [
    "offline",
    "streaming"
  ],
  "languages": [
    "en",
    "de",
    "fr",
    "es",
    "it",
    "pt",
    "nl",
    "pl",
    "ru",
    "uk",
    "cs",
    "ro",
    "hu",
    "sv",
    "da",
    "fi",
    "no",
    "el",
    "bg",
    "sk",
    "hr",
    "sr",
    "tr"
  ],
  "runtime": {
    "tags": [
      "gguf"
    ]
  },
  "capabilities": {
    "tts": [
      "long_form"
    ]
  },
  "options": {
    "request": [
      {
        "name": "voice_id",
        "type": "enum",
        "values": [
          "default",
          "clear",
          "english_female",
          "english_male"
        ],
        "description": "Bundled preset voice.",
        "default": "default",
        "required": false
      },
      {
        "name": "num_inference_steps",
        "type": "int",
        "description": "Diffusion steps per acoustic frame, from 1 to 1000; omitted uses the checkpoint's diffusion_head_config.ddpm_num_inference_steps.",
        "min": 1,
        "max": 1000,
        "default": 20,
        "required": false
      },
      {
        "name": "guidance_scale",
        "type": "float",
        "description": "Diffusion classifier-free guidance; 1 disables the unconditional pass. Any finite value is accepted; negative values enter CFG arithmetic.",
        "default": 3,
        "required": false
      },
      {
        "name": "max_tokens",
        "type": "int",
        "description": "Maximum autoregressive tokens per text chunk.",
        "min": 1,
        "default": 2048,
        "required": false
      },
      {
        "name": "do_sample",
        "type": "bool",
        "description": "Sample autoregressive speech control tokens instead of greedy decoding.",
        "default": false,
        "required": false
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "Autoregressive sampling temperature when do_sample is enabled. Any finite value is accepted; negative values disable sampling.",
        "default": 1,
        "required": false
      },
      {
        "name": "seed",
        "type": "int",
        "description": "Random seed; -1 chooses a random seed.",
        "min": -1,
        "default": 1234,
        "required": false
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Maximum Unicode codepoints per long-form text chunk.",
        "min": 1,
        "default": 300,
        "required": false
      },
      {
        "name": "text_chunk_mode",
        "type": "enum",
        "values": [
          "default",
          "tag_aware",
          "japanese",
          "endline"
        ],
        "description": "Framework long-form text splitting mode.",
        "default": "default",
        "required": false
      }
    ],
    "session": [
      {
        "name": "weight_type",
        "type": "enum",
        "values": [
          "native",
          "f32",
          "f16",
          "bf16",
          "q4_0",
          "q4_1",
          "q5_0",
          "q5_1",
          "q2_k",
          "q3_k",
          "q4_k",
          "q5_k",
          "q6_k",
          "q8_0",
          "nvfp4"
        ],
        "description": "Weight storage type; native preserves packaged tensor types.",
        "default": "native",
        "required": false
      }
    ],
    "load": []
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "kugelaudio_0_open_bf16",
      "display_name": "KugelAudio-0-Open BF16 GGUF",
      "default": false,
      "format": "gguf",
      "precision": "bf16",
      "target_directory": "KugelAudio-0-Open-GGUF",
      "files": [
        "KugelAudio-0-Open-GGUF/kugelaudio-0-open-bf16.gguf"
      ],
      "strip_prefix": "KugelAudio-0-Open-GGUF"
    },
    {
      "id": "kugelaudio_0_open_q8_0",
      "display_name": "KugelAudio-0-Open Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "KugelAudio-0-Open-GGUF",
      "files": [
        "KugelAudio-0-Open-GGUF/kugelaudio-0-open-q8_0.gguf"
      ],
      "strip_prefix": "KugelAudio-0-Open-GGUF"
    },
    {
      "id": "kugelaudio_0_open_q4_k",
      "display_name": "KugelAudio-0-Open Q4_K GGUF",
      "default": false,
      "format": "gguf",
      "precision": "q4_k",
      "target_directory": "KugelAudio-0-Open-GGUF",
      "files": [
        "KugelAudio-0-Open-GGUF/kugelaudio-0-open-q4_k.gguf"
      ],
      "strip_prefix": "KugelAudio-0-Open-GGUF"
    }
  ],
  "dependencies": [],
  "ui": {
    "recommended_package": "kugelaudio_0_open_q8_0",
    "default_voice": "default",
    "builtin_voices": [
      "default",
      "clear",
      "english_female",
      "english_male"
    ],
    "tags": [
      "TTS",
      "Stream",
      "GGUF"
    ],
    "docs": [
      "docs/models/kugelaudio.md",
      "docs/gguf.md"
    ]
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "tokenizer_json": "model:text_tokenizer/tokenizer.json",
        "tokenizer_config": "model:text_tokenizer/tokenizer_config.json"
      },
      "tensors": {
        "model_weights": "weights:"
      }
    }
  ]
}