Stabilization of low-light video

In one embodiment, a method includes determining a maximum exposure time and a first candidate exposure time for capturing one or more image frames from a video clip. The first candidate exposure time is based on an amount of motion of the computing device and a light level. If the first candidate exposure time is less than the maximum exposure time, then the method includes capturing the image frames with the first candidate exposure time. If the first candidate exposure time is greater than the maximum exposure time, then the method includes dividing the first candidate exposure time by an integer to determine a second candidate exposure time and capturing the image frames with the second candidate exposure time. A number of image frames equal to the integer are then combined.