news
VML South Africa develops Asili AI to combat bias in AI-generated African imagery
VML South Africa has launched Asili AI, an AI image generator trained to create authentic images of African people.
The tool addresses biases in existing AI generators that struggle with authentic representation of South African and African individuals.
Named after the Swahili word "Asili," meaning "origin" or "essence," it focuses on culturally significant imagery rooted in African heritage. The absence of authentic African imagery in stock photography has long been a challenge.
Matthew Arnold, chief innovation officer, VML South Africa: “As South Africans, we have always struggled to get decent stock photography featuring real South African people as opposed to stereotypical or Westernised/Americanised people.”
The explosion of AI has not addressed this issue, as the lack of authentic South African training material persists.
For creatives, this gap can lead to costly and time-consuming alternatives, as suitable imagery often needs to be shot from scratch. This can turn simple projects into expensive productions. To address this issue, VML developed its own solution.
Arnold: “We needed to create our own stock photography that allows for a much more authentic African feel.”
With AI advancing rapidly, incorporating it into the solution made sense. The team developed Asili AI, an image generator designed to produce culturally significant and authentic African imagery.
Asili is based on Stable Diffusion's Flux model and was meticulously trained by the team that will use it. Over six months, the internal team sourced and curated a large dataset of open-source, commercially ready images. The images were carefully labelled to capture nuances such as hair length and eye colour.
Arnold: “The biggest challenge was scale.”
Ensuring the dataset was extensive enough to train the model effectively was critical. Additionally, the team had to refine upscaling techniques to ensure images maintained crisp details, even for large formats like billboards.
Asili now produces high-quality, photorealistic images of African people in authentic settings.
By addressing a specific need, VML has developed a tool that enhances its work and tackles the representation gap in stock photography. For now, Asili is available exclusively to VML’s team and clients, offering a significant advantage in delivering culturally resonant visuals.
The tool will continue evolving, with a focus on expanding location-specific imagery for South Africa and other African countries.
Arnold: “Our main focus now is increasing the number of specific South African locations and starting to cater for other African countries.”
This ongoing refinement aims to provide a broader footprint of authentic African imagery.
