{
  "family": "miotts",
  "display_name": "MioTTS",
  "description": "Lightweight LLM-based English and Japanese TTS family built on MioCodec, supporting low-latency speech generation and zero-shot voice cloning from short reference audio.",
  "category": "tts",
  "status": "supported",
  "tasks": [
    "tts",
    "clone"
  ],
  "modes": [
    "offline"
  ],
  "languages": [
    "en",
    "ja"
  ],
  "capabilities": {
    "clone": [
      "speaker_reference"
    ]
  },
  "options": {
    "request": [
      {
        "name": "max_tokens",
        "type": "int",
        "description": "Maximum generated MioCodec tokens; default 700.",
        "required": false,
        "min": 1,
        "default": 700
      },
      {
        "name": "do_sample",
        "type": "bool",
        "description": "Enable stochastic token sampling; default comes from model generation config, falling back to true.",
        "required": false,
        "default": true
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "Token sampling temperature in [0, 2]; default comes from model generation config, falling back to 0.8.",
        "required": false,
        "min": 0.0,
        "max": 2.0,
        "default": 0.8
      },
      {
        "name": "top_k",
        "type": "int",
        "description": "Token top-k sampling limit; default comes from model generation config, falling back to 50.",
        "required": false,
        "min": 0,
        "default": 50
      },
      {
        "name": "top_p",
        "type": "float",
        "description": "Token nucleus sampling limit in [0, 1]; default comes from model generation config, falling back to 1.0.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 1.0
      },
      {
        "name": "repetition_penalty",
        "type": "float",
        "description": "Token repetition penalty in [1, 1.5]; default comes from model generation config, falling back to 1.0.",
        "required": false,
        "min": 1.0,
        "max": 1.5,
        "default": 1.0
      },
      {
        "name": "presence_penalty",
        "type": "float",
        "description": "Token presence penalty in [0, 1]; default 0.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 0.0
      },
      {
        "name": "frequency_penalty",
        "type": "float",
        "description": "Token frequency penalty in [0, 1]; default 0.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 0.0
      },
      {
        "name": "seed",
        "type": "int",
        "description": "Sampling seed; omitted value uses a random seed.",
        "required": false,
        "min": 0
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Per-request long-form text chunk size in Unicode codepoints; overrides session text_chunk_size. Default session value is 180.",
        "required": false,
        "min": 1
      },
      {
        "name": "best_of_n",
        "type": "int",
        "description": "Generate n candidates for this request; overrides session best_of_n_default and is capped by session best_of_n_max.",
        "required": false,
        "min": 1
      },
      {
        "name": "best_of_n_enabled",
        "type": "bool",
        "description": "Enable or disable best-of-N candidate selection for this request; overrides session best_of_n_enabled.",
        "required": false,
        "default": false
      },
      {
        "name": "best_of_n_language",
        "type": "enum",
        "description": "Language used when scoring best-of-N candidates for this request; overrides session best_of_n_language.",
        "preset": "best_of_n_language",
        "required": false,
        "default": "auto"
      }
    ],
    "session": [
      {
        "name": "codec_path",
        "type": "path",
        "description": "MioCodec model directory used by MioTTS acoustic decoding; default is the sibling MioCodec package path when present.",
        "required": false
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Default long-form text chunk size in Unicode codepoints; default 180.",
        "required": false,
        "min": 1,
        "default": 180
      },
      {
        "name": "best_of_n_enabled",
        "type": "bool",
        "description": "Enable best-of-N candidate selection by default.",
        "required": false,
        "default": false
      },
      {
        "name": "best_of_n_default",
        "type": "int",
        "description": "Default best-of-N candidate count; default 1.",
        "required": false,
        "min": 1,
        "default": 1
      },
      {
        "name": "best_of_n_max",
        "type": "int",
        "description": "Maximum best-of-N candidate count; default 8.",
        "required": false,
        "min": 1,
        "default": 8
      },
      {
        "name": "best_of_n_language",
        "type": "enum",
        "description": "Default language used when scoring best-of-N candidates.",
        "preset": "best_of_n_language",
        "required": false,
        "default": "auto"
      },
      {
        "name": "best_of_n_asr_path",
        "type": "path",
        "description": "Qwen3-ASR model directory used by best-of-N scoring; default is sibling Qwen3-ASR-0.6B when present.",
        "required": false
      },
      {
        "name": "weight_type",
        "type": "enum",
        "description": "MioTTS language-model matmul and WavLM weight storage type; default native.",
        "preset": "weight_type_full",
        "required": false,
        "default": "native"
      },
      {
        "name": "weight_context_mb",
        "type": "int",
        "description": "MioTTS language-model weight context size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      },
      {
        "name": "prefill_graph_arena_mb",
        "type": "int",
        "description": "MioTTS language-model prefill graph arena size in MiB; default 1024.",
        "required": false,
        "min": 1,
        "default": 1024
      },
      {
        "name": "decode_graph_arena_mb",
        "type": "int",
        "description": "MioTTS language-model decode graph arena size in MiB; default 1024.",
        "required": false,
        "min": 1,
        "default": 1024
      },
      {
        "name": "codec_weight_context_mb",
        "type": "int",
        "description": "MioCodec decoder weight context size in MiB; default 256.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "codec_constant_context_mb",
        "type": "int",
        "description": "MioCodec decoder constant tensor context size in MiB; default 256.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "global_graph_arena_mb",
        "type": "int",
        "description": "MioCodec global reference encoder graph arena size in MiB; default 256.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "wave_graph_arena_mb",
        "type": "int",
        "description": "MioCodec wave decoder graph arena size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      }
    ],
    "load": []
  },
  "runtime": {
    "tags": [
      "gguf"
    ]
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "generation_config": "model:generation_config.json",
        "tokenizer_config": "model:tokenizer_config.json",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt"
      },
      "optional_files": {
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "weights": "weights:"
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.json",
        "generation_config": "model:generation_config.json",
        "tokenizer_config": "model:tokenizer_config.json",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt"
      },
      "optional_files": {
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "weights": "model:model.safetensors"
      }
    }
  ],
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "miotts_1_7b_q8_0",
      "display_name": "MioTTS 1.7B Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "MioTTS-1.7B-GGUF",
      "files": [
        "MioTTS-1.7B-GGUF/miotts-1.7b-q8_0.gguf"
      ],
      "strip_prefix": "MioTTS-1.7B-GGUF"
    },
    {
      "id": "miotts_1_7b_bf16",
      "display_name": "MioTTS 1.7B BF16 GGUF",
      "format": "gguf",
      "precision": "bf16",
      "target_directory": "MioTTS-1.7B-GGUF",
      "files": [
        "MioTTS-1.7B-GGUF/miotts-1.7b-bf16.gguf"
      ],
      "strip_prefix": "MioTTS-1.7B-GGUF"
    },
    {
      "id": "miotts_1_7b_orig",
      "display_name": "MioTTS 1.7B Original-Dtype GGUF",
      "format": "gguf",
      "precision": "orig",
      "target_directory": "MioTTS-1.7B-GGUF",
      "files": [
        "MioTTS-1.7B-GGUF/miotts-1.7b-orig.gguf"
      ],
      "strip_prefix": "MioTTS-1.7B-GGUF"
    }
  ],
  "dependencies": [
    {
      "kind": "model",
      "family": "miocodec",
      "scope": "session",
      "option": "codec_path",
      "required": true
    },
    {
      "kind": "model",
      "family": "qwen3_asr",
      "scope": "session",
      "option": "best_of_n_asr_path",
      "required": false,
      "required_when": [
        {
          "scope": "session",
          "option_key": "miotts.best_of_n_enabled",
          "equals": true
        },
        {
          "scope": "request",
          "option_key": "best_of_n_enabled",
          "equals": true
        }
      ]
    }
  ]
}