Simultaneous measurements of marine snow (particles larger than 600 µm) morphologies, estimates of their in situ sinking speeds and midwater attenuation in export plumes were performed for the first time using a BGC-Argo float equipped with optical and imaging sensors. The float was deployed and recovered after one year drifting in the sluggish flow regime of the Angola basin. Six consecutive chlorophyll-a and particulate matter accumulation events were recorded at the surface, each followed by an export plume of sinking aggregates. Objects larger than 600 µm were classified using machine learning recognition and clustered into four morphological categories of marine aggregates. Plankton images were validated by an expert in a few broad categories. Results show that different types of aggregates were produced and exported from the different blooms. The different morphological categories of marine snow had different sinking speeds and attenuation for similar size indicating the effect of morphology on sinking speed. However, the typical size-to-sinking relationship for two of the categories and over the larger observed size range (100 µm-few mm) was also observed, indicating the importance of size for sinking. Surprisingly, calculated in situ sinking speeds were constantly in the lower range of known values usually assessed ex situ, suggesting a methodological effect which is discussed. Moving away from purely size-based velocity relationships and incorporating these additional morphological aggregates properties will help to improve mechanistic understanding of particle sinking and provide more accurate flux estimates. When used from autonomous platforms at high frequency, they will also provide increased spatio-temporal resolution for the observation of intermittent export events naturally occurring or induced by human activities associated with marine Carbon Dioxide Removal.