Soma Labels will enable the subjective labelling of body-focused data, including that generated by Soma Skins. Current labelling tools that contribute to training pipelines assume there are ‘correct’ labels; a view made explicit in work on robust classification in the face of ‘noisy labels’ which sees ‘true labels’ as being massively outnumbered by ‘incorrect’ ones. In contrast, Soma Labels will encourage its users to create, share and compare labels that are ambiguous, divergent and that celebrate outliers. These might represent personal feelings and opinions, express important qualities of physical experience from different disciplinary perspectives, or bring feminist, crip or queer critical perspectives to bear on data in the interests of inclusivity. The resulting datasets might be used to train deep networks or be visualised so that audiences can compare experiences while researchers can generate and map concepts.