voice_audio_multi/model_specs/audio8_tts.json

{
  "schema_version": 1,
  "family": "audio8_tts",
  "display_name": "Audio8 TTS Preview 0.6B",
  "description": "Audio8 TTS Preview 0.6B DualAR text-to-speech model with expressive multilingual speech across yue, zh, nl, en, fr, de, it, ja, ko, pl, and es, automatic language handling, and rapid voice cloning from short reference samples.",
  "category": "tts",
  "status": "community",
  "dependencies": [],
  "tasks": [
    "tts",
    "clone"
  ],
  "modes": [
    "offline",
    "streaming"
  ],
  "languages": [
    "auto",
    "yue",
    "zh",
    "nl",
    "en",
    "fr",
    "de",
    "it",
    "ja",
    "ko",
    "pl",
    "es"
  ],
  "capabilities": {
    "clone": [
      "speaker_reference"
    ]
  },
  "runtime": {
    "tags": [
      "gguf",
      "stream"
    ]
  },
  "options": {
    "request": [
      {
        "name": "reference_text",
        "type": "string",
        "description": "Reference transcript used with speaker reference audio.",
        "required": false
      },
      {
        "name": "multi_reference_cond",
        "type": "string",
        "description": "Ordered Audio8 TTS reference conditioning pairs as a JSON array: [{\"audio\":\"ref.wav\",\"text\":\"transcript\"}, ...].",
        "required": false
      },
      {
        "name": "max_tokens",
        "type": "int",
        "description": "Maximum semantic tokens to generate; default 1024. Zero selects the default.",
        "required": false,
        "min": 0,
        "default": 1024
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Maximum characters per generated text chunk; default 200.",
        "required": false,
        "min": 1,
        "default": 200
      },
      {
        "name": "text_chunk_mode",
        "type": "enum",
        "description": "Framework text chunking mode.",
        "preset": "text_chunk_mode_full",
        "required": false,
        "default": "word_budget"
      },
      {
        "name": "top_p",
        "type": "float",
        "description": "Top-p sampling value in (0, 1]; default 0.8.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 0.8
      },
      {
        "name": "top_k",
        "type": "int",
        "description": "Top-k sampling value; default 30.",
        "required": false,
        "min": 1,
        "default": 30
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "Semantic-token sampling temperature in (0, 2); default 0.8.",
        "required": false,
        "min": 0.0,
        "max": 2.0,
        "default": 0.8
      },
      {
        "name": "seed",
        "type": "int",
        "description": "Sampling seed for reproducible output; omitted uses a random seed.",
        "required": false,
        "min": 0
      },
      {
        "name": "language",
        "type": "string",
        "description": "Language hint for Chinese-script normalization; Cantonese/yue preserves Traditional characters. Inherits structured transcript or voice language when provided.",
        "required": false
      }
    ],
    "session": [
      {
        "name": "mem_saver",
        "type": "bool",
        "description": "Release cached AR runtime graphs after each request; default false.",
        "required": false,
        "default": false
      },
      {
        "name": "reference_cache_slots",
        "type": "int",
        "description": "Prepared reference-audio cache slots; default 1.",
        "required": false,
        "min": 0,
        "default": 1
      },
      {
        "name": "weight_type",
        "type": "enum",
        "description": "AR matmul weight storage type; default native.",
        "preset": "weight_type_full",
        "required": false,
        "default": "native"
      },
      {
        "name": "codec_weight_type",
        "type": "enum",
        "description": "Codec conv/matmul weight storage type; default native.",
        "preset": "weight_type_codec_q8",
        "required": false,
        "default": "native"
      },
      {
        "name": "ar_graph_arena_mb",
        "type": "int",
        "description": "AR runtime graph arena size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      },
      {
        "name": "ar_weight_context_mb",
        "type": "int",
        "description": "AR weight context size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      },
      {
        "name": "codec_graph_arena_mb",
        "type": "int",
        "description": "Codec encode/decode graph arena size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      },
      {
        "name": "codec_weight_context_mb",
        "type": "int",
        "description": "Codec weight context size in MiB; default 512.",
        "required": false,
        "min": 1,
        "default": 512
      }
    ],
    "load": []
  },
  "ui": {
    "recommended_package": "audio8_tts_preview_0_6b_q8_0",
    "tags": [
      "TTS",
      "Clone"
    ],
  "docs": [
    "docs/models/audio8_tts.md"
  ]
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "js-byte/Audio8-TTS-Preview-0.6b-GGUF",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "audio8_tts_preview_0_6b_q8_0",
      "display_name": "Audio8 TTS Preview 0.6B Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "Audio8-TTS-Preview-0.6B-GGUF",
      "files": [
        "audio8-tts-preview-0.6b-q8_0.gguf"
      ]
    }
  ],
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "tokenizer_config": "model:tokenizer_config.json",
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "model_weights": {
          "source": "weights:",
          "prefix": "model_weights"
        },
        "codec_weights": {
          "source": "weights:",
          "prefix": "codec_weights"
        }
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.json",
        "tokenizer_config": "model:tokenizer_config.json",
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "model_weights": "model:model.safetensors",
        "codec_weights": "model:codec.safetensors"
      }
    }
  ]
}