Dynamic hardware generation is a powerful technique thatcan substantially reduce both the required hardware resour-ces and the time needed to perform a calculation, reﬂectedin an improved functional density. This performance im-provement is a result of additional run-time optimizationsenabled by the knowledge of values at certain inputs at run-time. However, due to the large overhead conventional hard-ware generation tools incur, the usability of dynamic hard-ware generation is limited. We present a dual approach thatcombines compile-time generation of generic hardware andrun-time specialization. This drastically decreases the dy-namic generation overhead. Our approach is used for dy-namic generation of FIR ﬁlters and compared to a static anda conventional dynamic implementation. The experimentsclearly show that the dual approach improves the usabilityof dynamic hardware generation.