Real-time video can be efficiently and reliably streamed over wireless networks by fountain encoding a stream of video packets at the frame-level. Specifically, deep packet inspection may be performed on the stream of video packets to identify a set of video packets corresponding to an intra-coded frame (I-frame). The set of video packet may then be fountain encoded together as a data segment to produce a set of fountain encoded packets corresponding to the I-frame. Video packets corresponding to other frames (e.g., P-frames, etc.) may also be encoded on a frame-by-frame basis. This may allow a user equipment (UE) to begin displaying earlier frames in a group of pictures (GoP) while fountain encoded packets for subsequent frames are still being received, thereby reducing latency and improving the overall user experience.