{
  "id": "xclip-base-patch32",
  "identity": {
    "family": "xclip",
    "display_name": "X-CLIP base-patch32",
    "hub_id": "microsoft/xclip-base-patch32",
    "api_name": null,
    "revision": null,
    "variants": [
      "microsoft/xclip-base-patch32-16-frames"
    ],
    "license": "MIT",
    "commercial_use": true,
    "commercial_use_note": "MIT per the hub tag and repo.",
    "release_date": "2022-08",
    "paper_url": "https://arxiv.org/abs/2208.02816",
    "repo_url": "https://github.com/microsoft/VideoX/tree/master/X-CLIP"
  },
  "task": {
    "category": "embedding",
    "modalities_in": [
      "video",
      "text"
    ],
    "modalities_out": [
      "vector"
    ],
    "best_at": "Cheap video-text matching on short clips with a CLIP backbone; zero-shot action labels (Kinetics-trained).",
    "worst_at": "Trained on Kinetics-400 action clips: weak on general scenes, logos, people and speech; 8 frames per clip."
  },
  "io": {
    "input": {
      "video": {
        "frames": 8,
        "resolution": "224x224"
      },
      "text": {
        "max_tokens": 77
      },
      "prompts": "action phrases"
    },
    "output": {
      "vector": {
        "dims": 512,
        "distance": "cosine",
        "type": "dense"
      }
    }
  },
  "mixpeek_wiring": {
    "extractors": [],
    "feature_uris": [],
    "vector_indexes": [],
    "retriever_stages": [
      "feature_search"
    ],
    "composes_with": [
      "scene splits (transcoding/multimodal extractor) -> clip-level vectors -> feature_search with a text query"
    ]
  },
  "run": {
    "status": "plugin",
    "status_reason": "Not served. Safetensors with transformers XCLIPModel (no remote code); would run as a batch plugin over scene clips, with realtime text query embedding.",
    "serve_apps": [],
    "runtime": "transformers XCLIPModel + XCLIPProcessor",
    "weights_format": "safetensors",
    "trust_remote_code": false,
    "size_on_disk_mb": 786,
    "params_millions": 197,
    "preprocessing": "Sample 8 frames uniformly, resize and center-crop to 224, CLIP mean/std.",
    "postprocessing": "L2-normalize video and text embeddings.",
    "serving_shape": "both",
    "batch_size": null,
    "serve_batch": null,
    "scale_to_zero": {
      "min_replicas": 0,
      "max_replicas": 1,
      "idle_timeout_s": 600,
      "memory_gb": 2,
      "num_cpus": 2,
      "num_gpus": 0,
      "cold_start_s": {
        "warm_node_s": null,
        "new_node_s": null,
        "source": null
      },
      "cpu_fallback": true
    }
  },
  "cost": {
    "unmeasured": true,
    "measured": null,
    "note": "No committed measurement; plan row 3 measures it in the standalone image."
  },
  "evidence": {
    "unmeasured": true,
    "items": []
  },
  "refs": [
    "hf://microsoft/xclip-base-patch32@a2e27a78a2b5d802e894b8a1ef14f3a8ce490963"
  ],
  "formats": [
    {
      "ref": "hf://microsoft/xclip-base-patch32@a2e27a78a2b5d802e894b8a1ef14f3a8ce490963",
      "format": "safetensors",
      "file": "model.safetensors",
      "loader": "transformers / open_clip from_pretrained; Mixpeek Hub import (api/models/huggingface.py)",
      "supported_today": true
    },
    {
      "ref": "hf://microsoft/xclip-base-patch32@a2e27a78a2b5d802e894b8a1ef14f3a8ce490963",
      "format": "pytorch_bin",
      "file": "pytorch_model.bin",
      "loader": "torch.load (pickle); Mixpeek refuses pickle weights (shared/security/torch_pickle.py)",
      "supported_today": false
    }
  ],
  "use_cases": [
    "video-understanding",
    "action-recognition"
  ],
  "recommended": {
    "value": false,
    "for": [],
    "reason": null
  },
  "demos": [],
  "seo": {
    "title": "X-CLIP base-patch32: embedding model",
    "description": "X-CLIP base-patch32 is a video and text embedding model (MIT). Can run on Mixpeek as a custom plugin.",
    "keywords": [
      "X-CLIP base-patch32",
      "xclip",
      "embedding",
      "microsoft/xclip-base-patch32",
      "video-understanding",
      "action-recognition"
    ],
    "faqs": [
      {
        "q": "What is X-CLIP base-patch32 best at?",
        "a": "Cheap video-text matching on short clips with a CLIP backbone; zero-shot action labels (Kinetics-trained)."
      },
      {
        "q": "What are the limits of X-CLIP base-patch32?",
        "a": "Trained on Kinetics-400 action clips: weak on general scenes, logos, people and speech; 8 frames per clip."
      },
      {
        "q": "Can X-CLIP base-patch32 be used commercially?",
        "a": "Yes. MIT per the hub tag and repo."
      },
      {
        "q": "How does X-CLIP base-patch32 run on Mixpeek?",
        "a": "Not served. Safetensors with transformers XCLIPModel (no remote code); would run as a batch plugin over scene clips, with realtime text query embedding."
      },
      {
        "q": "What does X-CLIP base-patch32 output?",
        "a": "A 512-dimensional dense vector, compared by cosine similarity."
      }
    ]
  },
  "last_verified": "2026-11-05",
  "sources": [
    "https://huggingface.co/microsoft/xclip-base-patch32",
    "https://arxiv.org/abs/2208.02816",
    "https://huggingface.co/docs/transformers/model_doc/xclip"
  ],
  "registry": null,
  "templates": [],
  "usage": {
    "namespaces": null,
    "collections": null,
    "uncounted": true,
    "templates": 0,
    "counted_at": null,
    "source": null
  },
  "facets": {
    "status": "plugin",
    "category": "embedding",
    "family": "xclip",
    "license": "MIT",
    "commercial_use": true,
    "modalities_in": [
      "video",
      "text"
    ],
    "modalities_out": [
      "vector"
    ],
    "use_cases": [
      "video-understanding",
      "action-recognition"
    ],
    "recommended": false,
    "release_date": "2022-08",
    "params_millions": 197,
    "size_on_disk_mb": 786,
    "vector_dims": 512,
    "vector_distance": "cosine",
    "weights_format": "safetensors",
    "formats": [
      "pytorch_bin",
      "safetensors"
    ],
    "loadable_today": true,
    "trust_remote_code": false,
    "cpu_fallback": true,
    "num_gpus": 0,
    "memory_gb": 2,
    "min_replicas": 0,
    "serving_shape": "both",
    "served": false,
    "ref_schemes": [
      "hf"
    ],
    "has_evidence": false,
    "cost_measured": false,
    "template_count": 0,
    "usage_namespaces": null,
    "usage_collections": null,
    "usage_uncounted": true,
    "has_demo": false,
    "rank_score": 1
  },
  "rank_score": 1
}