{
  "camera_index": 0,
  "model_path": "models/dev-baseline.json",
  "confidence_threshold": 0.7,
  "hold_frames": 3,
  "tts_provider": "mock",
  "transcript_path": "outputs/demo-transcript.jsonl",
  "metrics_path": "outputs/eval-metrics.json",
  "dataset_path": "data/processed/sample_landmarks.jsonl",
  "clip_manifest_path": "data/processed/sample_sentence_clips.jsonl",
  "max_frames": 18,
  "frame_interval_seconds": 0.0,
  "use_camera": true,
  "model_mode": "word",
  "cnn_model_path": "models/cnn-baseline.keras",
  "cnn_frame_count": 16,
  "cnn_image_size": 160,
  "cnn_batch_size": 4,
  "cnn_epochs": 8,
  "vlm_provider": "mock",
  "vlm_model_id": "Qwen/Qwen2.5-VL-32B-Instruct-AWQ",
  "demo_sequence": [
    "HELLO",
    "YES",
    "NO",
    "THANK_YOU",
    "HELP",
    "STOP"
  ],
  "elevenlabs_api_key": null,
  "elevenlabs_voice_id": null
}