voice_audio_multi/model_specs/qwen3_forced_aligner.json

{
  "family": "qwen3_forced_aligner",
  "display_name": "Qwen3 Forced Aligner",
  "description": "Qwen3 non-autoregressive forced aligner that aligns transcript text to speech and returns word- or character-level timestamps across 11 supported languages.",
  "category": "speech_analysis",
  "status": "supported",
  "tasks": [
    "align"
  ],
  "modes": [
    "offline"
  ],
  "languages": [
    "zh",
    "en",
    "yue",
    "fr",
    "de",
    "it",
    "ja",
    "ko",
    "pt",
    "ru",
    "es"
  ],
  "capabilities": {
    "align": [
      "word_timestamps"
    ]
  },
  "options": {
    "request": [
      {
        "name": "clamp_timestamps_to_audio",
        "type": "bool",
        "description": "Opt-in guard for word timestamp output: keep repaired word spans inside the local audio input. Defaults to false to preserve existing timestamp repair behavior.",
        "required": false,
        "default": false
      }
    ],
    "session": [
      {
        "name": "weight_type",
        "type": "enum",
        "description": "Shared thinker matmul storage; does not override audio encoder storage.",
        "required": false,
        "preset": "weight_type_full",
        "default": "native"
      },
      {
        "name": "audio_encoder_weight_type",
        "type": "enum",
        "description": "Audio encoder matmul storage.",
        "required": false,
        "values": [
          "native",
          "f32",
          "f16"
        ],
        "default": "native"
      },
      {
        "name": "thinker_weight_type",
        "type": "enum",
        "description": "Thinker matmul storage; inherits weight_type when omitted.",
        "required": false,
        "preset": "weight_type_full"
      },
      {
        "name": "audio_encoder_graph_arena_mb",
        "type": "int",
        "description": "Audio encoder graph metadata arena. Size in MiB; must be positive.",
        "required": false,
        "min": 1,
        "default": 128
      },
      {
        "name": "thinker_prefill_graph_arena_mb",
        "type": "int",
        "description": "Thinker prefill graph metadata arena. Size in MiB; must be positive.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "thinker_decode_graph_arena_mb",
        "type": "int",
        "description": "Thinker cached-step graph metadata arena. Size in MiB; must be positive.",
        "required": false,
        "min": 1,
        "default": 256
      },
      {
        "name": "thinker_weight_context_mb",
        "type": "int",
        "description": "Thinker weight metadata context. Size in MiB; must be positive.",
        "required": false,
        "min": 1,
        "default": 64
      }
    ],
    "load": []
  },
  "runtime": {
    "tags": [
      "gguf"
    ]
  },
  "ui": {
    "recommended_package": "qwen3_forced_aligner_0_6b_q8_0",
    "tags": [
      "Align",
      "GGUF"
    ],
    "docs": [
      "docs/models/qwen3_forced_aligner.md",
      "docs/gguf.md"
    ]
  },
  "package_defaults": {
    "download": {
      "kind": "huggingface_snapshot",
      "repo": "audio-cpp/audio.cpp-gguf",
      "revision": "main",
      "gated": false
    }
  },
  "packages": [
    {
      "id": "qwen3_forced_aligner_0_6b_q8_0",
      "display_name": "Qwen3 Forced Aligner 0.6B Q8_0 GGUF",
      "default": true,
      "format": "gguf",
      "precision": "q8_0",
      "target_directory": "Qwen3-ForcedAligner-0.6B-GGUF",
      "files": [
        "Qwen3-ForcedAligner-0.6B-GGUF/qwen3-forced-aligner-0.6b-q8_0.gguf"
      ],
      "strip_prefix": "Qwen3-ForcedAligner-0.6B-GGUF"
    },
    {
      "id": "qwen3_forced_aligner_0_6b_f16",
      "display_name": "Qwen3 Forced Aligner 0.6B F16 GGUF",
      "format": "gguf",
      "precision": "f16",
      "target_directory": "Qwen3-ForcedAligner-0.6B-GGUF",
      "files": [
        "Qwen3-ForcedAligner-0.6B-GGUF/qwen3-forced-aligner-0.6b-f16.gguf"
      ],
      "strip_prefix": "Qwen3-ForcedAligner-0.6B-GGUF"
    },
    {
      "id": "qwen3_forced_aligner_0_6b_safetensors",
      "display_name": "Qwen3 Forced Aligner 0.6B Safetensors",
      "format": "safetensors",
      "precision": "native",
      "target_directory": "Qwen3-ForcedAligner-0.6B",
      "files": [
        "config.json",
        "generation_config.json",
        "model.safetensors",
        "preprocessor_config.json",
        "tokenizer_config.json"
      ],
      "download": {
        "kind": "huggingface_snapshot",
        "repo": "Qwen/Qwen3-ForcedAligner-0.6B"
      }
    }
  ],
  "sources": [
    {
      "format": "gguf",
      "roots": {
        "model": ".",
        "weights": "$gguf"
      },
      "files": {
        "config": "model:config.json",
        "generation_config": "model:generation_config.json",
        "tokenizer_config": "model:tokenizer_config.json"
      },
      "optional_files": {
        "preprocessor_config": "model:preprocessor_config.json",
        "processor_config": "model:processor_config.json",
        "chat_template": "model:chat_template.json",
        "chat_template_jinja": "model:chat_template.jinja",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt",
        "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"
      },
      "optional_files": {
        "preprocessor_config": "model:preprocessor_config.json",
        "processor_config": "model:processor_config.json",
        "chat_template": "model:chat_template.json",
        "chat_template_jinja": "model:chat_template.jinja",
        "vocab": "model:vocab.json",
        "merges": "model:merges.txt",
        "tokenizer_json": "model:tokenizer.json"
      },
      "tensors": {
        "weights": "model:model.safetensors"
      }
    }
  ]
}