A speaker, having a firmer structure, includes a metal holder, a magnetic circuit unit including a first diaphragm, and a vibration unit including a clamp plate. The clamp plate includes long-axis edges and short-axis edges. The holder includes long edges, short edges, and corner portions. The short edge includes a first surface facing away from the first diaphragm. The short-axis edge abuts against the first surface. The short-axis edge includes a second surface facing the short edge. The corner portion includes a third surface facing away from the first diaphragm. The holder further includes a protrusion extending from the first surface in a direction facing away from the first diaphragm. The clamp plate further includes an avoiding portion formed by recessing from the second surface in a direction facing away from the short edge. The protrusion is received in and connected to the avoiding portion.