Listen here to the article.
This is the second chapter of an in-depth study on AI that we decided to do after the second edition of Digital Content Factory Revolution, the event held last June at our headquarters in Affi (VR) and launched in June 2023 to share annual news and reflections on the revolution that AI is bringing to the processes of content production and distribution in the fashion industry.
In the first chapter (read more here), we saw some insights of a more general and methodical nature, in this second chapter we will instead see some more punctual application areas that emerged during the day, in particular the areas and workflows for which we have already envisaged AI integration in the core engine of our Chalco BrandLife solution. We will distinguish between two major areas: the world of images and visual content on the one hand, and the world of product information and textual content on the other.
1_ Images and Visuals: Post-production and Automatic Adaptations
The first area refers to the integration in Chalco of a series of AI services capable of automating some of the most costly and time-consuming post-production actions. These include the automatic generation of masks and paths for the automatic contouring (1) of even very complex images and a first level of beauty correction on complexion and colour homogeneity, texture and correction of imperfections.
Automatic contouring (1)
The second important area is the AI functionality related to automatic adaptations generated from a master image according to channel, context or platform specifications. We are talking about extremely useful functionalities that can already provide significant time and cost savings on the process. Among these, the automatic management of positional adaptations (2) of the product allows, for example, the consistent management of the virtual shelf in PLP e-commerce and thus the correct positioning of different products with different dimensions.
Positional adaptations (2)
Another very useful service already available in Chalco is the automatic cut (3) of wearable images to preset levels on the model’s face – useful e.g. in the context of rights management of casting in different channels and geographies.
Automatic cut (3)
Child in its turn of the automatic contouring mentioned above, the background change service (4): from the simplest colour modification treatments up to the recontextualisation of the subject (still life or worn) in real or simulated contexts completely different from the shooting background.
Last among the services we mention here, the generation or modification of shadows in both cases of still-life and wearable shots: think for instance of the automatic generation of shadows with different positions and depths in a still life shot of accessories or the possibility of managing the intensity and depth of contact shadows downstream of a wearable shot.
Background change (4)
2_ Product information and texts: auto-tagging, descriptions and translations
The second area refers to the contribution of AI in the generation of product metadata, information and text from images or 3D renderings. The initial step in this area is represented by what is commonly referred to as auto-tagging, i.e. the ability of the AI service to identify a set of features from the product image and then compile a list of tags or metadata useful to describe the product.
In this case, the service is able to identify even complex characteristics such as the product category, formal characteristics (type of collar, sleeve length, RGB colour code, patterns, logos or designs to name but a few in the fashion sphere), up to belonging to a reference style, creative theme or collection.
The fact that the first service was able to identify the peculiar characteristics of an object, in turn puts the second service in a position to produce a product description that can be adapted and customised for different uses or targets – one thinks, for example, of the possibility of generating different descriptions for use in a B2B digital showroom or an e-commerce PDP, up to a declination of the content for different social platforms.
It almost goes without saying at this point that a third AI service can be used to generate translations and localisations depending on the geographical or cultural context of the market.
All the services mentioned above can also benefit from the availability of pre-existing information: think for instance of the information already available in ERP and PLM systems which, together with product images, represent an extremely useful resource for the generation of ever richer and more accurate content.
