{
  "model": "Phantom-Wan-1.3B",
  "framework": "mlx",
  "task": "subject-to-video",
  "dtype": "bfloat16",
  "dim": 1536,
  "num_layers": 30,
  "num_heads": 12,
  "in_dim": 16,
  "patch_size": [
    1,
    2,
    2
  ],
  "cross_attention_dim": 4096,
  "vae_z_dim": 16,
  "vae_stride": [
    4,
    8,
    8
  ],
  "quantization_4bit": {
    "group_size": 64,
    "bits": 4,
    "skip": [
      "patch_embedding_proj",
      "text_embedding",
      "time_embedding",
      "time_projection",
      "head"
    ],
    "cosine_vs_bf16": 0.99633
  },
  "sample": {
    "shift": 5.0,
    "steps": 50,
    "guide_img": 5.0,
    "guide_text": 7.5,
    "fps": 16
  }
}