Give your apps the gift of sight.

Mixpeek is flexible vision understanding infrastructure that's built to scale with you. Use our APIs to index, search, classify, generate and analyze videos and images for your most ambitious applications.

Get Started Read Docs

Even bring your own database Read more

There's a method for everything

Embed Scene

Embed scene extracts key information from video frames, providing a rich understanding of the visual content.


{
  "scene": {
    "embedding": [0.1, 0.2, 0.3, 0.4],
    "objects": ["car", "tree", "person"],
    "actions": ["driving", "walking"],
    "setting": "urban street",
    "time_of_day": "daytime",
    "weather": "sunny"
  }
}

Face detection identifies and analyzes human faces in images or video frames.


{
  "faces": [
    {
      "bounding_box": [100, 50, 200, 150],
      "confidence": 0.98,
      "landmarks": {
        "left_eye": [120, 80],
        "right_eye": [180, 80],
        "nose": [150, 100],
        "mouth_left": [130, 130],
        "mouth_right": [170, 130]
      },
      "emotions": {
        "happy": 0.7,
        "neutral": 0.3
      }
    }
  ]
}

Audio transcription converts spoken words in audio files to written text.


{
  "transcription": [
    {
      "start_time": "00:00:01",
      "end_time": "00:00:05",
      "speaker": "Speaker 1",
      "text": "Welcome to our video on AI-powered video analysis."
    },
    {
      "start_time": "00:00:06",
      "end_time": "00:00:10",
      "speaker": "Speaker 2",
      "text": "Today, we'll explore how machine learning can extract insights from video content."
    }
  ]
}

Text reading extracts and recognizes text present in images or video frames.


{
  "text_regions": [
    {
      "bounding_box": [50, 100, 300, 150],
      "text": "AI-Powered Video Analysis",
      "confidence": 0.95
    },
    {
      "bounding_box": [75, 200, 275, 250],
      "text": "Extracting Insights",
      "confidence": 0.92
    }
  ]
}

Activity description provides a detailed analysis of actions and events occurring in the video.


{
  "activities": [
    {
      "timestamp": "00:00:05",
      "description": "A person is jogging in a park",
      "confidence": 0.95,
      "objects": ["person", "trees", "path"],
      "actions": ["jogging", "moving"]
    },
    {
      "timestamp": "00:00:15",
      "description": "A dog is playing fetch with its owner",
      "confidence": 0.92,
      "objects": ["person", "dog", "ball"],
      "actions": ["throwing", "running", "catching"]
    }
  ]
}

...or use everything at once

You can choose to use each method individually or just index the entire video for end-to-end search. They can come from a live camera feed or object storage like AWS S3.

Integrate

Connect your data sources securely to Mixpeek's processing pipeline.

Integration Docs

Index

Leverage 100s of models to pull out data from your files or feeds.

Indexing Docs

Analyze

Leverage your newly-structured data to build apps powered by previously unaccessible data.

Use Case Docs

              
mixpeek.search("person jogging in park with dog")

{
  "results": [
    {
      "start_time": 0,
      "end_time": 5,
      "embedding": [0.1, 0.2, 0.3, 0.4],
      "faces": ["face.jpg"],
      "transcription": {
        "text": "It's a beautiful day for a jog in the park.",
        "speaker": "Narrator"
      },
      "text": [
        {
          "text": "Park Entrance",
          "bounding_box": [50, 100, 300, 150],
          "confidence": 0.95
        }
      ],
      "descriptions": {
        "description": "A person is jogging on a path in a sunny park",
        "confidence": 0.92
      }
    }
  ]
}