Beam-shape transformers are used in high-efficiency projectors to match round source beams to rectangular LCD targets. Two common alternative approaches use (1) light- pipes and (2) lenslet arrays. Both approaches outperform the simple overfilling used in simpler systems. We review the two approaches conceptually and develop simple theoretical models which elucidate each approach's trade-offs between uniformity, efficiency, etendue, and compactness. For the light-pipe approach, we develop a detailed analytical theory for idealized, uniform sources, and we compare the predictions to ray-trace results. Finally, we compare the two approaches and find that the light-pipe approach offers similar performance with higher compactness. The lens array approach, however, may be favored in complex systems where multiple elements already result in long optical trains, or in systems where etendue conservation is not a priority.