Mobile imaging devices face significant variability in image quality due to differences in hardware, environmental conditions,and device-specific characteristics, posing challenges for consistent photo enhancement. Traditional methods often rely ondevice-specific calibrations or computationally intensive algorithms, restricting their usability in environments with limitedresources. To overcome these constraints, we propose the Multimodal Diffusion Forecaster, a lightweight and device-agnosticframework for photo enhancement. This framework incorporates three core modules: the Manifold Constrained DiffusionFilter, the Agent-driven Reflectance Mapper, and the Probabilistic Noise Reconstruction Unit. These modules collectivelyensure robust image enhancement while maintaining computational efficiency, making the approach suitable for mobile devices.The model leverages multimodal data representations and diffusion-based mechanisms to achieve high-quality enhancementacross diverse device profiles. Furthermore, an uncertainty-aware reconstruction strategy is introduced to quantify andmitigate uncertainties caused by device-specific variations and environmental inconsistencies, improving the reliability of theenhancement process. Experimental evaluations demonstrate that the proposed framework outperforms existing methods interms of enhancement quality, structural fidelity, and device-agnostic generalization. The results show a significant improvementin image quality, with an average enhancement accuracy increase of 12% compared to baseline methods, while maintaininglow computational overhead. This work provides a practical solution for consistent and efficient photo enhancement, enablingits deployment in resource-constrained mobile environments and advancing mobile imaging technology.