Our AI in a nutshell

Frameright AI is the world's first Dynamic Visual Machine Learning model. Trained on a dataset of over 1 million professionally cropped images, it suggests optimal composition for your story, style, application and display size.

From here we can start training a model that knows your style. This means you will know what your AI is basing it’s decisions on.

Transparency and unique style makes you visually strong.

Values and vision

Our style-aware and customer-specific models help you develop and protect your unique look and stand out from the competition.

We are constantly refining our technology and training the AI on different types of images. Football images, for example, are different from portraits.

And you always have the final say on how your images appear: we believe in an “augmented” intelligence approach, fusing the power of AI with human intelligence and creativity.

This approach also reflects our belief in diversity, transparency and integrity: we don’t believe in “one size fits all” in any sense of the term. We are committed to removing AI bias.

How it works

Composing images is much more than having an object in the frame. This is why we go beyond object detection. In order to understand what kind of frame is aesthetically pleasing we based our model on a set of images that were all cropped in several shapes by professional image editors. 

Images in different categories e.g. portraits or football images require different kind of approach and this is what we are currently getting better at. At the end of the day creative decisions are always complex and that is why it is our priority to grant you complete freedom to adjust the suggestions for your final edits.

We use a deep learning approach built on top of our custom convolutional neural networks to improve our machine learning models.  This means that the image goes through several layers of processing, examining it from multiple perspectives and thanks to some mathematical magic, our network gets more accurate over time.

Cutomize for style and transparency

Aesthetics are subjective. We understand that every organisation and individual has their own idea of what makes a “good photograph”. We grant you the power to define your own style that evolves with you.

We’ll deploy our in-house model on your company’s dataset to understand your framing and style preferences. We’ll then fine-tune our hyper parameters and create an AI that understands your style and aesthetics. 

This way you can always know what kind of data your model is built on. Knowing the ingredients of your model is one of the minimum requirements in fighting against biased AI.