The transmission of live video over noisy channels requires very low end-to-end delay. Although automatic repeat re-quest ensures lossless transmission, its usefulness to live video streaming is restricted to short connections because of the unbounded retransmission latency. An alternative is to use forward error correction (FEC). Since finding an op-timal error protection strategy can be time expensive, FEC systems are commonly designed for the worst case con-dition of the channel, which limits the end-to-end perfor-mance. We study the suitability of two scalable FEC-based systems to the transmission of live video over packet net-works. The first one uses Reed-Solomon codes and is ap-propriate for the Internet. The second one uses a product channel code and is appropriate for wireless channels. We show how fast and robust transmission can be achieved by exploiting a parametric model for the distortion-rate curve of the source coder and by using fast joint source-channel allocation algorithms. Experimental results for the 3D set partitioning in hierarchical tree video coder show that the systems have good reconstruction quality even in severe chan-nel conditions. Finally, we compare the performance of the systems to the state-of-the-art for video transmission over the Internet.