Techniques for dynamically grouping items include determining context associated with each group comprising one or more items; and determining current context at a device. In response to determining that prompted prompt is presented for input to indicate an item, a suggested group is determined based on the current context at the device and relevance of a context associated with the suggested group. The suggested group is presented. In some embodiments, prompted prompt is presented to name a suggested group; and, in response, data indicating the suggested group is stored as a permanent group associated with the name. In some embodiments, relevance is determined for each item of a group. A group avatar is generated based on M images associated with the items that have the most relevance.