Reebok perfectsplit
A web-based project to generate custom video-gifs from #perfectsplit images

Nuestras contradicciones son lo que realmente nos definen. Coincidiendo con la campaña #perfectsplit de Reebok, realizamos junto con la agencia Soon in Tokio un generador automático de gifs. A partir de dos imágenes de una persona, dos versiones de una misma personalidad, la cara A < prudente, calmada, inteligente, el día... > y la cara B < activa, atrevida, directa, la noche...>, desarrollamos una web que creaba una composición única para cada usuario mostrando el contraste entre ambas.

Fotografiamos a Alba Paul & Diego Berrueco dos influencers del mundo de la moda y las tendencias, y con sus dos caras activamos la acción. Una experiencia digital con conectividad en redes y abierta a todos los públicos.


Technical Discussion

This simple project was very challenging in terms of technology.

The web was powered by a backend application that used computer vision to recognise the face contour and the different elements: eyes, eyebrows... of the user images. This allow to create meaningful gifs in which the graphic elements where positioned in the correct place not such randomly. The service was programmed in Openframeworks run in a headless GPU server-cloud. It could render a unique video for every input in less than two seconds creating a customise experience for every user.

The main challenge was to build a platform that could create gifs with the appropriate format to share natively on different social networks: Instagram, twitter and Facebook.

To achieve this, the gifs where automatically transformed to mp4 video format, of three second length and low bandwidth, and they were sent by email to the user. Then he or she could share the video-gif on Instagram directly from the mobile phone with no further transformations. Furthermore a copy of the gif was stored in the cloud, which allowed the application to create Facebook a twitter native sharing cards. So instead of sharing just a link, users could share the moving-images in their social networks which improves enormously the engagement of the action.


Some tests

Developed by: Espadaysantacruz Studio
Agency: Soon in Tokio
Client: Reebok
Powered by: Ruby on Rails, NodeJs & Openframeworks

Sept - 2016