top of page



Automatic Story Discovery


  • Finds in which country a photo belongs to with no network access required;

  • Infers home country and frequent locations through machine learning;

  • Finds Travel scenarios based on these two reference area types;

  • Defines small moments based on space and time gap between photos;

  • Classifies each photo individually based on moment, travel, frequent location and country;

  • Transforms user camera rolls into memories with different scales of aggregation for unusual scenarios like a travel and daily life events.

foto selection.png

Best Photo Selection

  • Computes a global ranking value based on content analysis and quality assessment; 

  • Similarity based on content, feature extraction and feature matching;

  • Storytelling relevant photo curation;

  • Aesthetics assessment through Deep Learning;

  • Quality metrics based on sharpness and exposure;

  • Photo global Aesthetics through Deep Learning;

  • Computes an individual ranking index for each photo based on all variables previously described;

  • Elastic curation targeting a specific number of photos, photo book pages or video seconds that keeps the story structure. For example, the tree created as the skeleton of the narrative is used to propagate representation time down tree until no more representation time is available. After that, photos are selected for each chapter or sub-chapter based on similarity and ranking.

deep learning.png

Photo Content Analysis

  • Filters images in or out based on content through Deep Learning. For example, a model can be used to filter out documents, business cards and whiteboards or a model can be used to find photos with sunsets;

  • Face detection and quality analysis;

  • Content detection to be used in photo ranking flow. For example, if a photo has a young baby, it can rank up;

  • Photo similarity analysis based on content through Deep Learning and by finding common points between images through feature extraction and matching;

  • Photo global Aesthetics through Deep Learning.


Contextual Storytelling


  • Either user selected or automatic discovered memories can be enriched with maps and naming through the narrative found based on space and time variability of the photos;

  • Variable level of information based on a predefined number of pages for a photo book or target number of seconds for a video story. For example, a video with 60 seconds can show maps zooming in into sub-chapter regions and a video with 30 seconds can only represent main chapters;

  • Contextual data like maps and labels based on individual photo classification carried out on “Automatic Stories”. For example, if all photos in the set were classified as travel, a contextual map can be added at the beginning of the book or video that includes both home and travel geographic areas;

  • Social Activity that can be matched in space and time with the story.


"Elastic" Storytelling


  • Stories can be curated to variable levels of detail;

  • Compresses or enlarges a story for a target number of photos, pages or seconds;

  • Stories are organised in a hierarchy of chapters and sub-chapters. As an example a story in a 12-page book will focus on the main chapters but the same story in a 40-page book will include more details and sub-chapters;

  • Organizes either an automatic discovered memory or a manually selected group of photos in a hierarchy of chapters and sub-chapters based on space and time variability of input set through unsupervised machine learning. For example, this can be used to shrink or enlarge a story for predefined target of pages to match minimum and maximum of your printing process constraints.


Fast, On-Device Execution


  • Runs 100% on the device;

  • Takes advantage dedicated HW for AI computation;

  • No cloud infrastructure required;

  • Guaranteed user privacy.

deep learning.png

Deep Learning


  • Pluggable neural network models for app size control and specificity;

  • Content based models dynamically deployed without app update;

  • Takes advantage of OS built-in models to avoid adding unnecessary bytes.


Real-Time video rendering engine


  • On-device, real-time video rendering leveraging dedicated HW units;

  • Rendering is carried out in html 5 making it portable to browsers;

  • Dynamic aspect ratio and resolution;

  • Themes can be updated dynamically from your server for app size management.


Small SDK binaries

  • Photo curation and storytelling SDK is independent from SDK for video rendering to avoid adding more bytes than those that are strictly necessary for your app;

  • Video rendering SDK itself adds less than 1 MB to your app and Photo book SDK adds less than 5 MB.

Anchor 1
Anchor 2
Anchor 3
Anchor 4
Anchor 5
Anchor 6
Anchor 7
Anchor 8
bottom of page