We present an interactive technique that can render aliasing-free and coherent word clouds. Words that are not legible are first identified and appropriately adjusted in size to remove aliasing. The adjusted words are either placed in a local optimal position without overlapping the existing words, or filtered out for effective utilization of screen space. The layout of the refined word cloud is then compressed by a novel force-directed model. We demonstrate the effectiveness and usefulness of our technique with an application in semantic zooming of word clouds.