Leveraging in-situ synchrotron x-ray imaging measurements of vapor depression in laser-based additive manufacturing, we develop an image-based multiphysics model, which promises to evaluate laser energy absorptivity directly using x-ray images as input. A deep learning algorithm is used to automatically extract the liquid-gas interface from the x-ray images. A ray-tracing method is then used to predict the laser absorptivity based on the detected liquid-gas interface. The predicted results are compared with real-time energy absorption measurements using in-situ integrating sphere radiometry. The developed model is beneficial to generate absorptivity data for understanding the effect of processing on energy absorption mechanisms. Moreover, the developed model can be regarded as a high-fidelity heat source model. It can be combined with other computational models for accurate simulations of melt pool dynamics, defect formation, and solidification microstructure in the additive manufacture of materials.