reflexion.ai is a content collaboration, AI, and storyboarding platform. It has 3 modules :

  1. Collaboration & Review

    This module helps companies manage their content on the platform and collaborate with other teams.

  2. Artificial Intelligence

    Every frame of the content is tagged by vision models and the entire content is tagged using audio & speech models.

  3. Storyboarding

    It is possible to search relevant sections of the content using the AI-based smart search, and drag-drop them to create a quick storyboard.

Building Blocks

  1. Available Platforms

    reflexion.ai is available on Web, IOS and Android

  2. Software as a Service

    Option to register on the platform followed by a review process followed by access to reflexion.ai.

  3. Model Performance

    Proprietary vision & audio/speech models optimized for Nvidia CUDA cores for high performance.

  4. Application Security

    Best-in-class cloud security protocols on Azure cloud. Encrypted communication through SSL.

  5. Content Privacy

    Option of hosting content on Azure private cloud. Watermark feature for downloading content.

  6. Content Delivery Network

    Azure CDN used for faster streaming, faster data upload, lower network latency & packet loss.

  7. Storage

    Highly secure, flexible & limitless storage along with proxy low-resolution content creation on the server.

Computer Vision

  1. Face Recognition

    Our vision models are trained to identify 1750 actors of Indian origin. We are scaling up to international figures as well. All unrecognized faces are clustered and shown separately which can be named by the user.

  2. Emotion Detection

    Our vision models are trained to identify up to 12 emotions. For every face, the model computes the percentage of each emotion. We have tried to cover extreme emotions such as sad & crying, and happy & laughing, etc.

  3. Object Identification

    Our vision models are trained to identify 500+ objects classified into 30 odd categories such as weapons, personal items, wearables, animals, food items, musical instruments, furniture, sports equipments, house items, etc.

  4. Action Identification

    Our vision models are trained to identify up to 400 actions such as playing certain sports, doing certain dance, eating certain items, hugging/kissing, etc.

  5. Place Identification

    Our vision models are trained to identify whether the shot is outdoor or indoor and detects the type of place such as coffee shop, store, hospital, museum, office, playground, etc.

  6. Object Character Recognition

    The character of any text written anywhere in the frame is recognized and stored as data. This can be used subsequently to search across the content.

  7. Detecting Compliance

    Our vision models are trained to identify smoking scenes, drinking scenes, violence and % skin exposure (nudity). These can be used by networks to ensure compliance messages are in place.

  8. Classifying Scenes & Shots

    Our vision models can identify different scenes. Further, within each scene, the different shots can be identified. Also, the type of camera shot (close-up, extra wide, etc) are also identified.

Audio & Speech

  1. Transcription & Translation

    We use Azure Cognitive APIs to get the transcript on the cleaned vocals and also translate into other languages.

  2. Sub-Titles Creation

    Our speech models create the closed caption / sub-titles file for the content. There are options to edit, download, upload the sub-titles file..

  3. Speaker Identification

    Our speech models are trained to identify the speaker amongst all the trained actors of Indian origin..

  4. Speaker Tone Identification

    Our speech models are trained to identify the tone of the speaker from the 12 trained tones/emotions.

  5. Urban Sounds Detection

    Our audio models are trained to detect 250 urban sounds such as dog barking, car honking, etc.

  6. Background Score Generation

    Based on background sounds, the background score is generated which can be used to identify climax scenes & songs.

  7. Music Genre Identification

    Our audio models are trained to identify the genre of music such as pop, classical, metal, instrumental, etc.

  8. Audio Fingerprinting

    We create an audio fingerprint of all uploaded content which can be used to identify the source of any uploaded content.