voice_audio_multi/model_specs/xtts_v2.json

{
  "schema_version": 1,
  "family": "xtts_v2",
  "display_name": "XTTS v2",
  "description": "Coqui XTTS v2 cross-lingual text-to-speech with voice cloning: GPT-2 mel-code generator, perceiver conditioning resampler, HiFi-GAN decoder, and speaker encoder.",
  "category": "tts",
  "status": "community",
  "tasks": [
    "tts"
  ],
  "modes": [
    "offline"
  ],
  "languages": [
    "en",
    "es",
    "fr",
    "de",
    "it",
    "pt",
    "pl",
    "tr",
    "ru",
    "nl",
    "cs",
    "ar",
    "zh-cn",
    "hu",
    "ko",
    "ja",
    "hi"
  ],
  "capabilities": {
    "tts": [
      "speaker_reference"
    ]
  },
  "options": {
    "request": [
      {
        "name": "language",
        "type": "string",
        "description": "Source language for text conditioning (e.g. 'en'); also selects tokenizer normalization.",
        "required": false,
        "default": "en"
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "AR sampling temperature; official default 0.75.",
        "required": false,
        "min": 0.0,
        "default": 0.75
      },
      {
        "name": "top_k",
        "type": "int",
        "description": "AR top-k sampling; official default 50.",
        "required": false,
        "min": 0,
        "default": 50
      },
      {
        "name": "top_p",
        "type": "float",
        "description": "AR nucleus sampling threshold; official default 0.85.",
        "required": false,
        "min": 0.0,
        "max": 1.0,
        "default": 0.85
      },
      {
        "name": "repetition_penalty",
        "type": "float",
        "description": "AR repetition penalty; official default 10.0.",
        "required": false,
        "min": 1.0,
        "default": 10.0
      },
      {
        "name": "max_mel_tokens",
        "type": "int",
        "description": "Maximum mel codes to generate; official default 605.",
        "required": false,
        "min": 1,
        "default": 605
      },
      {
        "name": "seed",
        "type": "int",
        "description": "Sampling seed; omitted selects a random seed.",
        "required": false,
        "min": 0,
        "max": 4294967295
      },
      {
        "name": "text_chunk_size",
        "type": "int",
        "description": "Maximum Unicode codepoints per framework text chunk; default 256.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "text_chunk_mode",
        "type": "enum",
        "description": "Framework text chunking mode.",
        "preset": "text_chunk_mode_full",
        "required": false,
        "default": "word_budget"
      }
    ],
    "session": [
      {
        "name": "conditionals_cache_slots",
        "type": "int",
        "description": "Prepared voice-conditioning cache slots; 0 disables caching.",
        "required": false,
        "min": 0,
        "default": 1
      }
    ],
    "load": []
  },
  "runtime": {
    "tags": [
      "gguf"
    ]
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "vocoder": "model:xtts-v2-vocoder-f16.gguf"
      },
      "tensors": {
        "gpt_weights": {
          "source": "weights:"
        },
        "vocoder_weights": {
          "source": "model:xtts-v2-vocoder-f16.gguf"
        }
      }
    }
  ],
  "packages": [
    {
      "id": "xtts_v2_gguf_q8_0",
      "display_name": "XTTS v2 GGUF (Q8_0)",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "XTTS-v2-Q8",
      "files": [
        "q8_0/xtts-v2-gpt-q8_0.gguf",
        "q8_0/xtts-v2-vocoder-f16.gguf",
        "q8_0/mel_norms.bin"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "onicubag/xTTSv2-GGUF"
      },
      "strip_prefix": "q8_0"
    },
    {
      "id": "xtts_v2_gguf",
      "display_name": "XTTS v2 GGUF (Original)",
      "default": false,
      "format": "gguf",
      "precision": "orig",
      "target_directory": "XTTS-v2",
      "files": [
        "xtts-v2-gpt.gguf",
        "xtts-v2-vocoder-f16.gguf",
        "mel_norms.bin"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "onicubag/xTTSv2-GGUF"
      }
    },
    {
      "id": "xtts_v2_gguf_q4_k_m",
      "display_name": "XTTS v2 GGUF (Q4_K_M)",
      "default": false,
      "format": "gguf",
      "precision": "q4_k_m",
      "target_directory": "XTTS-v2-Q4_K_M",
      "files": [
        "q4_k_m/xtts-v2-gpt-q4_k_m.gguf",
        "q4_k_m/xtts-v2-vocoder-f16.gguf",
        "q4_k_m/mel_norms.bin"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "onicubag/xTTSv2-GGUF"
      },
      "strip_prefix": "q4_k_m"
    }
  ],
  "dependencies": [],
  "ui": {
    "tags": [
      "TTS",
      "Clone",
      "GGUF"
    ],
    "docs": [
      "docs/community_models/xtts_v2.md"
    ],
    "recommended_package": "xtts_v2_gguf_q8_0"
  }
}