Image matting is a visual technology that can be applied to robots for visual perception and environmental understanding. Natural image matting algorithms strive to predict the alpha matte with the assistance of trimaps. Nonetheless, the generation of trimaps poses a significant challenge, constraining their applicability in various practical contexts. This is because trimaps need to be manually created, and pixel-level foreground/background/unknown annotation involves high labor costs.To address the common practical problem, we introduce TAMatting, a general matting framework that converts trimap-guided methods to mask-guided approaches without additional training. It can be applied to most matting methods and attains comparable performance, making high-quality matting more accessible. The cornerstone of TAMatting lies in its trimap adaptation, an innovative procedure that utilizes a coarse mask as a guide to produce a precise trimap. Subsequently, a pre-trained matting model predicts alpha mattes with the generated trimaps. Experimental results demonstrate that our TAMatting is widely applicable in addressing practical challenges. Additionally, we offer qualitative results on practical applications, showing the significant impact and potential of our approach.