Non-terrestrial networks (NTNs), particularly satellite networks, are key factors for ubiquitous connectivity in today's 5G terrestrial networks (TNs) and future 6G networks. They can bring great benefits, but also great challenges. The direct-to-cell connectivity (D2C) use case, i.e., direct connectivity from satellites to common terrestrial low-gain user terminals (UTs), such as handheld or IoT devices, is currently gaining strong interest. This paper focuses on innovative distributed implementations of the space segment using swarms of multiple small platforms, each embedding one or a subset of radiating elements. They are organized with regular or irregular geometries and greater distance between radiating elements, creating large virtual antenna array apertures. The use of small satellites promises reduced production and launch costs, while the distributed nature of the system provides attractive features, such as scalability and fault tolerance. This paper analyzes the multi-beam performance of two distributed implementations, comparing the results with a single platform implementation. Distributed implementations present a better performance in terms of average summed throughput, user throughput, and throughput density. An analysis of worst-case user throughput reveals that distributed implementations based on irregular geometries capable of mitigating grating lobes provide slightly lower average performance of measured key performance indicators (KPIs), but greater fairness than the implementation based on regular geometries. Swarm-based antenna arrays appear as a promising solution for the D2C connectivity use case.