To realize multiple-input multiple-output (MIMO) with low cost and power consumption, the use of low-resolution digital-to-analog converters (DAC) is drawing significant interest. While previous studies on massive MIMO systems with low resolution DACs have focused on optimizing precoding, this approach faces the limitation that the coarse resolution at the transmitter makes it difficult to generate the transmit signal close enough to the desired one when a small scale MIMO transmitter is considered as in IoT networks. To address this problem, we propose a new transmission framework that jointly performs modulation and beamforming. With the proposed scheme, the message symbol is directly mapped into a feasible transmit vector. We also develop a channel-adaptive mapping function for the proposed framework that minimizes error probability. Simulation results show that the proposed scheme can improve the error performance