Service features
Media AI video analysis service that finds all the meaning in the video
Metadata and personal information search
Through AI video analysis, it provides comprehensive keywords such as objects, behaviors, time, and space. It is also easy to check where a specific person appears in the video.
Voice recognition and automatic subtitling features
It uses a multimodal analysis engine to automatically recognize not only images but also dialogue and generate a script. It distinguishes each speaker's voice and integrates it with personal information.
Scene segmentation with context understanding
MAIU analyzes video in "sections" rather than separate frames, greatly improving search speed and inference efficiency. This provides even richer insights based on meaningful scenes.
Service advantages
Structured insights
MAIU extracts key insights based on the results of scene segmentation and provides them in a structured way so that they can be understood at a glance.
Metadata and personal information search
In addition to objects, you can search for the desired section using various basic keywords such as actions such as running and clapping, or time and place. You can also easily check where a specific person appears in the video.
Free and flexible use
Data can be freely modified by person labeling and dialogue modification. Users can easily find the keywords they are looking for by entering them directly as text.
Service diagram
MAIU diagram
MAIU stores and uses low-resolution files converted to the minimum size required for video analysis. The converted files are used only for video analysis and not for training to upgrade the engine.
Learn MoreService utilization method
Archive search
You can easily browse and use the database through the MAIU's index DB.
Content reproduction
You can use the scene search feature to reproduce content that meets your needs, such as highlight clips.
Recommendation association service
Detailed features
Section-by-section insight extraction
It provides the results of understanding and analyzing video in meaningful scene units rather than in general frame units.
It provides comprehensive vision and audio information for individual scenes to support quick and efficient video understanding.
Metadata extraction and search
You can quickly find the desired segment by selecting keywords related to more than 500 objects, behaviors,
time and space, and shooting techniques that are provided by default, or by entering natural language directly into the prompt.
In addition, you can search for and extract the desired scene segment from a large number of indexes stored in the index DB
on a single page.
Person recognition and DB support
When analyzing video, the information on the person you want to search for can be directly built and stored in the database and used in the analysis process.
Extensive insight modification and utilization
Customization of detailed metadata such as person, keyword, speaker, and script is possible.
It provides download of a JSON file containing metadata extracted after video analysis.
This can be used for various scenarios, such as easy search of existing archived content or finding editing points.
Supported upload file specifications (table)
The following is a guide to the file formats and limits suitable for video analysis.
Input | |
|---|---|
Container Format | MP4, MKV, MOV, WEBM |
Resolution | 720p (1280 x 720) |
Audio Language | Korean, English |
File Size Limits | Up to 2 GB* |
File Duration Limits | 6 hours |
FPS | 10 fps recommended |
* The list of video file formats that can be added will be continuously improved during the CBT period.