{
  "schema_version": 1,
  "family": "personaplex",
  "display_name": "PersonaPlex",
  "description": "PersonaPlex is a Moshi-style full-duplex speech-to-speech conversational model with Mimi audio tokenization, packaged speaker/persona prompts, streaming audio input, text generation, and generated assistant speech.",
  "category": "tts",
  "status": "supported",
  "tasks": [
    "s2s"
  ],
  "modes": [
    "offline",
    "streaming"
  ],
  "languages": [
    "en"
  ],
  "runtime": {
    "tags": [
      "gguf",
      "stream"
    ]
  },
  "capabilities": {
    "s2s": [
      "speaker_reference"
    ]
  },
  "options": {
    "request": [
      {
        "name": "voice_id",
        "type": "enum",
        "description": "Packaged PersonaPlex voice prompt id such as NATF2 or NATM1. If omitted, NATF2 is used.",
        "values": [
          "NATF0",
          "NATF1",
          "NATF2",
          "NATF3",
          "NATM0",
          "NATM1",
          "NATM2",
          "NATM3",
          "VARF0",
          "VARF1",
          "VARF2",
          "VARF3",
          "VARF4",
          "VARM0",
          "VARM1",
          "VARM2",
          "VARM3",
          "VARM4"
        ],
        "required": false,
        "default": "NATF2"
      },
      {
        "name": "system_prompt",
        "type": "string",
        "description": "System persona prompt. Plain text is wrapped with the required <system> tags.",
        "required": false
      },
      {
        "name": "temperature",
        "type": "float",
        "description": "Audio token sampling temperature; must be nonnegative, and positive when do_sample=true. Ignored when do_sample=false; default 0.8.",
        "required": false,
        "min": 0.0,
        "default": 0.8
      },
      {
        "name": "text_temperature",
        "type": "float",
        "description": "Text token sampling temperature; must be nonnegative, and positive when do_sample=true. Ignored when do_sample=false; default follows temperature.",
        "required": false,
        "min": 0.0
      },
      {
        "name": "top_k",
        "type": "int",
        "description": "Audio token top-k sampling limit; default 250.",
        "required": false,
        "min": 0,
        "default": 250
      },
      {
        "name": "text_top_k",
        "type": "int",
        "description": "Text token top-k sampling limit; default follows top_k.",
        "required": false,
        "min": 0
      },
      {
        "name": "do_sample",
        "type": "bool",
        "description": "Enable stochastic token sampling; default true. Set false for greedy decoding.",
        "required": false,
        "default": true
      },
      {
        "name": "seed",
        "type": "int",
        "description": "uint32 seed. Session RNG starts at 42424242; explicit values reseed it, while omission on later conversations continues RNG state.",
        "required": false,
        "min": 0,
        "max": 4294967295
      }
    ],
    "session": [
      {
        "name": "graph_arena_mb",
        "type": "int",
        "description": "Reusable ggml graph arena size in MiB for PersonaPlex LM, depformer, and Mimi graphs; default 1024.",
        "required": false,
        "min": 1,
        "default": 1024
      },
      {
        "name": "lm_weight_context_mb",
        "type": "int",
        "description": "Main LM weight metadata arena size in MiB; default 64.",
        "required": false,
        "min": 1,
        "default": 64
      },
      {
        "name": "depformer_weight_context_mb",
        "type": "int",
        "description": "Depth transformer weight metadata arena size in MiB; default 64.",
        "required": false,
        "min": 1,
        "default": 64
      },
      {
        "name": "mimi_weight_context_mb",
        "type": "int",
        "description": "Mimi codec weight metadata arena size in MiB; default 64.",
        "required": false,
        "min": 1,
        "default": 64
      },
      {
        "name": "weight_type",
        "type": "enum",
        "description": "LM and Mimi matmul weight storage type; default native.",
        "preset": "weight_type_full",
        "required": false,
        "default": "native"
      }
    ],
    "load": []
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "personaplex_7b_v1_q4_k",
      "display_name": "PersonaPlex 7B v1 Q4_K GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q4_k",
      "target_directory": "PersonaPlex-GGUF",
      "files": [
        "PersonaPlex-GGUF/personaplex-7b-v1-q4_k.gguf"
      ],
      "strip_prefix": "PersonaPlex-GGUF"
    },
    {
      "id": "personaplex_7b_v1_q8_0",
      "display_name": "PersonaPlex 7B v1 Q8_0 GGUF",
      "default": false,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "PersonaPlex-GGUF",
      "files": [
        "PersonaPlex-GGUF/personaplex-7b-v1-q8_0.gguf"
      ],
      "strip_prefix": "PersonaPlex-GGUF"
    }
  ],
  "dependencies": [],
  "ui": {
    "recommended_package": "personaplex_7b_v1_q4_k",
    "tags": [
      "TTS",
      "Stream",
      "GGUF"
    ],
    "docs": [
      "docs/models/personaplex.md",
      "docs/tts.md",
      "docs/gguf.md"
    ]
  },
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "tokenizer_model": "model:tokenizer_spm_32k_3.model",
        "voice_NATF0": "model:voices_safetensors/NATF0.safetensors",
        "voice_NATF1": "model:voices_safetensors/NATF1.safetensors",
        "voice_NATF2": "model:voices_safetensors/NATF2.safetensors",
        "voice_NATF3": "model:voices_safetensors/NATF3.safetensors",
        "voice_NATM0": "model:voices_safetensors/NATM0.safetensors",
        "voice_NATM1": "model:voices_safetensors/NATM1.safetensors",
        "voice_NATM2": "model:voices_safetensors/NATM2.safetensors",
        "voice_NATM3": "model:voices_safetensors/NATM3.safetensors",
        "voice_VARF0": "model:voices_safetensors/VARF0.safetensors",
        "voice_VARF1": "model:voices_safetensors/VARF1.safetensors",
        "voice_VARF2": "model:voices_safetensors/VARF2.safetensors",
        "voice_VARF3": "model:voices_safetensors/VARF3.safetensors",
        "voice_VARF4": "model:voices_safetensors/VARF4.safetensors",
        "voice_VARM0": "model:voices_safetensors/VARM0.safetensors",
        "voice_VARM1": "model:voices_safetensors/VARM1.safetensors",
        "voice_VARM2": "model:voices_safetensors/VARM2.safetensors",
        "voice_VARM3": "model:voices_safetensors/VARM3.safetensors",
        "voice_VARM4": "model:voices_safetensors/VARM4.safetensors"
      },
      "tensors": {
        "lm_weights": {
          "source": "weights:",
          "prefix": "lm"
        },
        "mimi_weights": {
          "source": "weights:",
          "prefix": "mimi"
        }
      }
    },
    {
      "format": "safetensors",
      "roots": {
        "model": "."
      },
      "files": {
        "config": "model:config.json",
        "tokenizer_model": "model:tokenizer_spm_32k_3.model",
        "voice_NATF0": "model:voices_safetensors/NATF0.safetensors",
        "voice_NATF1": "model:voices_safetensors/NATF1.safetensors",
        "voice_NATF2": "model:voices_safetensors/NATF2.safetensors",
        "voice_NATF3": "model:voices_safetensors/NATF3.safetensors",
        "voice_NATM0": "model:voices_safetensors/NATM0.safetensors",
        "voice_NATM1": "model:voices_safetensors/NATM1.safetensors",
        "voice_NATM2": "model:voices_safetensors/NATM2.safetensors",
        "voice_NATM3": "model:voices_safetensors/NATM3.safetensors",
        "voice_VARF0": "model:voices_safetensors/VARF0.safetensors",
        "voice_VARF1": "model:voices_safetensors/VARF1.safetensors",
        "voice_VARF2": "model:voices_safetensors/VARF2.safetensors",
        "voice_VARF3": "model:voices_safetensors/VARF3.safetensors",
        "voice_VARF4": "model:voices_safetensors/VARF4.safetensors",
        "voice_VARM0": "model:voices_safetensors/VARM0.safetensors",
        "voice_VARM1": "model:voices_safetensors/VARM1.safetensors",
        "voice_VARM2": "model:voices_safetensors/VARM2.safetensors",
        "voice_VARM3": "model:voices_safetensors/VARM3.safetensors",
        "voice_VARM4": "model:voices_safetensors/VARM4.safetensors"
      },
      "tensors": {
        "lm_weights": "model:model.safetensors",
        "mimi_weights": "model:tokenizer-e351c8d8-checkpoint125.safetensors"
      }
    }
  ]
}