Marine stock assessments or biodiversity monitoring studies, which historically relied on extractive techniques (e.g., trawl or grab surveys), are being progressively replaced by non-extractive approaches. For instance, species abundance indices can be calculated using data obtained from high-definition underwater cameras that enable to identify taxa at low taxonomical level. In biodiversity studies, environmental DNA (eDNA) has proven to be a useful tool for characterising fish species richness. However, several marine phyla remain poorly represented in reference gene databases or release limited amounts of DNA, restricting their detection. The absence of amplification of some invertebrate taxa might also reflect primer bias. We here explore and compare the performance of eDNA and image data in describing the marine communities of several sites in the Bay of Biscay. This was achieved by deploying a remotely operated vehicle to both record images and collect seawater samples. A total of 88 taxa were identified from the eDNA samples and 121 taxa from the images. For both methods, the best characterised phylum was Chordata, with 29 and 27 Actinopterygii species detected using image versus eDNA, respectively. Neither Bryozoa nor Cnidaria was detected in the eDNA samples while the phyla were easily identifiable by imagery. Similarly, Asteroidea (Echinodermata) and Cephalopoda (Mollusca) were scarcely detected in the eDNA samples but present on the images, while Annelida were mostly identified by eDNA (18 taxa vs 7 taxa from imagery). The complementary community descriptions we highlight from these two methods therefore advocate for using both eDNA and imagery in tandem in order to capture the macroscopic biodiversity of bentho-demersal marine communities.