{
  "format": "ba-config-1",
  "ba_version": "0.1.0",
  "base_model": {
    "repo_id": "openbmb/VoxCPM2",
    "revision": "7936080b41163874f38610f46f438e2c0757f9fd",
    "licence": "Apache-2.0",
    "weights_modified": false,
    "files": [
      "config.json",
      "model.safetensors",
      "audiovae.pth",
      "tokenizer.json",
      "tokenizer_config.json",
      "special_tokens_map.json",
      "tokenization_voxcpm2.py"
    ]
  },
  "asr": {
    "repo_id": "openai/whisper-large-v3-turbo",
    "revision": "41f01f3fe87f28c78e2fbf8b568835947dd65ed9",
    "licence": "MIT",
    "use": "transcribes the reference recording when no transcript is given"
  },
  "watermark": {
    "repo_id": "facebook/audioseal",
    "revision": "3c19eba53390776cf2cc9ed5f6c9ac67ce72ecba",
    "licence": "MIT",
    "generator": "generator_base.pth",
    "detector": "detector_base.pth",
    "message": "BA",
    "message_bits": [
      0,
      1,
      0,
      0,
      0,
      0,
      1,
      0,
      0,
      1,
      0,
      0,
      0,
      0,
      0,
      1
    ]
  },
  "languages": {
    "arz": "Egyptian Arabic",
    "en": "English",
    "fr": "French (beta)"
  },
  "defaults": {
    "per_language": {
      "arz": {
        "cfg_value": 2.0,
        "inference_timesteps": 10,
        "max_chunk_chars": 1000,
        "chars_per_sec_band": [
          4.006,
          19.261
        ],
        "latin_words": "arabic"
      },
      "en": {
        "cfg_value": 2.0,
        "inference_timesteps": 10,
        "max_chunk_chars": 1000,
        "chars_per_sec_band": [
          4.462,
          22.389
        ]
      },
      "fr": {
        "cfg_value": 2.0,
        "inference_timesteps": 10,
        "max_chunk_chars": 1000,
        "chars_per_sec_band": [
          6.55,
          31.638
        ]
      }
    },
    "seed": 1234,
    "max_text_chars": 5000,
    "max_retries_per_chunk": 2,
    "pauses_s": {
      "sentence": 0.3,
      "clause": 0.15,
      "paragraph": 0.6
    },
    "output_lufs": -20.0,
    "reference": {
      "sample_rate": 16000,
      "min_speech_s": 3.0,
      "target_s": [
        6.0,
        15.0
      ],
      "noise_reject_db": 25.9,
      "noise_warn_db": 32.7,
      "clip_max_ratio": 0.00625,
      "lufs": -20.0
    }
  },
  "read_at_load": [
    "defaults.per_language.*.cfg_value",
    "defaults.per_language.*.inference_timesteps",
    "defaults.per_language.*.max_chunk_chars",
    "asr.repo_id",
    "asr.revision"
  ]
}