The disclosed technology relates to adaptively encoding a video stream based on viewer engagement, content schedule and/or content characteristics. A system routes a video stream from a signal extractor to a first virtual encoder. The first virtual encoder is instantiated on a cloud platform and provides a first video output at a first bitrate. A second virtual encoder is instantiated on the cloud platform in response to a change in viewer count. The system routes the video stream from the signal extractor to the second virtual encoder. The second virtual encoder provides a second video output at a second bitrate. The second bitrate is different from the first bitrate.