Bass guitar plays a preeminent role in Western Popular Music, having to convey both rhythmic and harmonic information. In this paper, we propose a transformer-based decoder model suggesting idiomatic bass guitar tablatures, conditioned on a rhythm guitar track. Using the DadaGP dataset and its tokenization scheme, we introduce a pipeline to identify rhythm guitar tracks and extract them along with the matching bass guitar tablature. The model is trained on over 100'000 16-bars rhythm guitar excerpts and generate relevant bass lines suggestions. We qualitatively analyze the generated bass tablatures and identify the key qualities and drawbacks of the system, like its ability to follow harmony, and rhythm whilst generating performable and natural hand movements for bass players, but with a tendency to copy the rhythm guitar too closely. All code and pretrained models are made publicly available, along with a demonstration tool.