{
  "family": "index_tts2",
  "display_name": "IndexTTS2",
  "description": "Zero-shot TTS system for Chinese and English speech synthesis with voice cloning, emotion-speaker decoupling, text or audio emotion control, and explicit duration control.",
  "category": "tts",
  "status": "supported",
  "tasks": [
    "tts",
    "clone"
  ],
  "modes": [
    "offline"
  ],
  "languages": [
    "zh",
    "en"
  ],
  "capabilities": {
    "tts": [
      "emotion_control"
    ],
    "clone": [
      "speaker_reference",
      "emotion_control"
    ]
  },
  "options": {
    "request": [
      {
        "name": "emotion_alpha",
        "type": "float",
        "description": "Blend strength for explicit emotion conditioning; default 1.0.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 1.0
      },
      {
        "name": "emotion_vector",
        "type": "float_list",
        "description": "Eight-value explicit emotion vector.",
        "required": false
      },
      {
        "name": "use_emotion_text",
        "type": "bool",
        "description": "Infer emotion from text instead of reference audio.",
        "required": false,
        "default": false
      },
      {
        "name": "emotion_text",
        "type": "string",
        "description": "Text used when emotion-text conditioning is enabled.",
        "required": false
      },
      {
        "name": "use_random_emotion",
        "type": "bool",
        "description": "Use random emotion weights in the emotion mixer.",
        "required": false,
        "default": false
      },
      {
        "name": "interval_silence_ms",
        "type": "int",
        "description": "Silence inserted between generated text chunks in milliseconds; default 200.",
        "required": false,
        "min": 0,
        "default": 200
      },
      {
        "name": "text_chunk_mode",
        "type": "enum",
        "description": "Framework text chunking mode used for long-form text; default word_budget.",
        "preset": "text_chunk_mode_full",
        "required": false,
        "default": "word_budget"
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Maximum characters per framework text chunk; omitted lets the tokenizer segment by its 120-token budget.",
        "required": false,
        "min": 1
      },
      {
        "name": "max_tokens",
        "type": "int",
        "description": "Maximum generated mel tokens; default 1500.",
        "required": false,
        "min": 1,
        "default": 1500
      },
      {
        "name": "do_sample",
        "type": "bool",
        "description": "Enable GPT sampling; default true.",
        "required": false,
        "default": true
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "GPT sampling temperature; must be positive when set; default 0.8.",
        "required": false,
        "min": 0.0,
        "default": 0.8
      },
      {
        "name": "top_k",
        "type": "int",
        "description": "GPT top-k sampling limit; default 30.",
        "required": false,
        "min": 1,
        "default": 30
      },
      {
        "name": "top_p",
        "type": "float",
        "description": "GPT nucleus sampling limit in (0, 1]; default 0.8.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 0.8
      },
      {
        "name": "length_penalty",
        "type": "float",
        "description": "GPT beam-search length penalty; default 0.0.",
        "required": false,
        "min": 0.0,
        "default": 0.0
      },
      {
        "name": "num_beams",
        "type": "int",
        "description": "GPT beam count; default 3.",
        "required": false,
        "min": 1,
        "default": 3
      },
      {
        "name": "repetition_penalty",
        "type": "float",
        "description": "GPT repetition penalty; must be positive when set; default 10.0.",
        "required": false,
        "min": 0.0,
        "default": 10.0
      },
      {
        "name": "seed",
        "type": "int",
        "description": "Generation seed for reproducible output; omitted uses a random seed.",
        "required": false,
        "min": 0
      }
    ],
    "session": [
      {
        "name": "weight_type",
        "type": "enum",
        "description": "Matmul weight storage type; default native.",
        "preset": "weight_type_full",
        "required": false,
        "default": "native"
      },
      {
        "name": "conv_weight_type",
        "type": "enum",
        "description": "Convolution weight storage type; default native.",
        "preset": "weight_type_conv",
        "required": false,
        "default": "native"
      },
      {
        "name": "gpt_graph_arena_mb",
        "type": "int",
        "description": "GPT graph arena size in MiB; default 2048.",
        "required": false,
        "min": 0,
        "default": 2048
      },
      {
        "name": "s2mel_graph_arena_mb",
        "type": "int",
        "description": "S2Mel graph arena size in MiB; default 2048.",
        "required": false,
        "min": 0,
        "default": 2048
      },
      {
        "name": "reference_graph_arena_mb",
        "type": "int",
        "description": "Reference encoder and codec graph arena size in MiB; default 512.",
        "required": false,
        "min": 0,
        "default": 512
      },
      {
        "name": "emotion_text_prefill_graph_arena_mb",
        "type": "int",
        "description": "Emotion-text prefill graph arena size in MiB; default 2048.",
        "required": false,
        "min": 0,
        "default": 2048
      },
      {
        "name": "emotion_text_decode_graph_arena_mb",
        "type": "int",
        "description": "Emotion-text cached-step graph arena size in MiB; default 512.",
        "required": false,
        "min": 0,
        "default": 512
      },
      {
        "name": "emotion_text_max_tokens",
        "type": "int",
        "description": "Maximum generated tokens for emotion-text classification; default 256.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "weight_context_mb",
        "type": "int",
        "description": "Shared weight context size in MiB; default 4096.",
        "required": false,
        "min": 0,
        "default": 4096
      },
      {
        "name": "mem_saver",
        "type": "bool",
        "description": "Release staged reference and conditioning graphs after request phases; default false.",
        "required": false,
        "default": false
      },
      {
        "name": "speaker_cache_slots",
        "type": "int",
        "description": "Prepared speaker-reference cache slots; default 1.",
        "required": false,
        "min": 0,
        "default": 1
      },
      {
        "name": "emotion_cache_slots",
        "type": "int",
        "description": "Prepared emotion-reference cache slots; default 1.",
        "required": false,
        "min": 0,
        "default": 1
      },
      {
        "name": "emotion_text_cache_slots",
        "type": "int",
        "description": "Emotion-text weight cache slots; default 1.",
        "required": false,
        "min": 0,
        "default": 1
      }
    ],
    "load": []
  },
  "runtime": {
    "tags": [
      "gguf"
    ]
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.yaml",
        "bpe": "model:bpe.model",
        "wav2vec2bert_config": "model:w2v-bert-2.0/config.json",
        "wav2vec2bert_preprocessor_config": "model:w2v-bert-2.0/preprocessor_config.json",
        "bigvgan_config": "model:bigvgan/config.json",
        "qwen_emotion_config": "model:qwen0.6bemo4-merge/config.json",
        "qwen_emotion_generation_config": "model:qwen0.6bemo4-merge/generation_config.json",
        "qwen_emotion_tokenizer": "model:qwen0.6bemo4-merge/tokenizer.json",
        "qwen_emotion_tokenizer_config": "model:qwen0.6bemo4-merge/tokenizer_config.json",
        "qwen_emotion_vocab": "model:qwen0.6bemo4-merge/vocab.json",
        "qwen_emotion_merges": "model:qwen0.6bemo4-merge/merges.txt"
      },
      "tensors": {
        "gpt": {
          "source": "weights:",
          "prefix": "gpt"
        },
        "s2mel": {
          "source": "weights:",
          "prefix": "s2mel"
        },
        "speaker_matrix": {
          "source": "weights:",
          "prefix": "speaker_matrix"
        },
        "emotion_matrix": {
          "source": "weights:",
          "prefix": "emotion_matrix"
        },
        "wav2vec2bert_stats": {
          "source": "weights:",
          "prefix": "wav2vec2bert_stats"
        },
        "wav2vec2bert": {
          "source": "weights:",
          "prefix": "wav2vec2bert"
        },
        "semantic_codec": {
          "source": "weights:",
          "prefix": "semantic_codec"
        },
        "campplus": {
          "source": "weights:",
          "prefix": "campplus"
        },
        "bigvgan": {
          "source": "weights:",
          "prefix": "bigvgan"
        },
        "qwen_emotion": {
          "source": "weights:",
          "prefix": "qwen_emotion"
        }
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.yaml",
        "bpe": "model:bpe.model",
        "wav2vec2bert_config": "model:w2v-bert-2.0/config.json",
        "wav2vec2bert_preprocessor_config": "model:w2v-bert-2.0/preprocessor_config.json",
        "bigvgan_config": "model:bigvgan/config.json",
        "qwen_emotion_config": "model:qwen0.6bemo4-merge/config.json",
        "qwen_emotion_generation_config": "model:qwen0.6bemo4-merge/generation_config.json",
        "qwen_emotion_tokenizer": "model:qwen0.6bemo4-merge/tokenizer.json",
        "qwen_emotion_tokenizer_config": "model:qwen0.6bemo4-merge/tokenizer_config.json",
        "qwen_emotion_vocab": "model:qwen0.6bemo4-merge/vocab.json",
        "qwen_emotion_merges": "model:qwen0.6bemo4-merge/merges.txt"
      },
      "tensors": {
        "gpt": "model:gpt.safetensors",
        "s2mel": "model:s2mel.safetensors",
        "speaker_matrix": "model:feat1.safetensors",
        "emotion_matrix": "model:feat2.safetensors",
        "wav2vec2bert_stats": "model:wav2vec2bert_stats.safetensors",
        "wav2vec2bert": "model:w2v-bert-2.0/model.safetensors",
        "semantic_codec": "model:semantic_codec_model.safetensors",
        "campplus": "model:campplus.safetensors",
        "bigvgan": "model:bigvgan/model.safetensors",
        "qwen_emotion": "model:qwen0.6bemo4-merge/model.safetensors"
      }
    }
  ],
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "index_tts2_q8_0",
      "display_name": "IndexTTS2 Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "IndexTTS2-GGUF",
      "files": [
        "IndexTTS2-GGUF/index-tts2-q8_0.gguf"
      ],
      "strip_prefix": "IndexTTS2-GGUF"
    },
    {
      "id": "index_tts2_f16",
      "display_name": "IndexTTS2 F16 GGUF",
      "format": "gguf",
      "precision": "f16",
      "target_directory": "IndexTTS2-GGUF",
      "files": [
        "IndexTTS2-GGUF/index-tts2-f16.gguf"
      ],
      "strip_prefix": "IndexTTS2-GGUF"
    },
    {
      "id": "index_tts2_orig",
      "display_name": "IndexTTS2 Original-Dtype GGUF",
      "format": "gguf",
      "precision": "orig",
      "target_directory": "IndexTTS2-GGUF",
      "files": [
        "IndexTTS2-GGUF/index-tts2-orig.gguf"
      ],
      "strip_prefix": "IndexTTS2-GGUF"
    },
    {
      "id": "index_tts2_safetensors",
      "display_name": "IndexTTS2 Safetensors",
      "format": "safetensors",
      "precision": "native",
      "target_directory": "IndexTTS-2",
      "files": [
        "config.yaml",
        "bpe.model",
        "gpt.safetensors"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "mlx-community/index-tts2-mlx"
      }
    }
  ]
}