🗳️ Requests

TBA. When running (Transcoding) Orchestrators, the lpData folder often fills up with temp files over time. Need to check the cleanup logic and perhaps clean all temp files on Orch start.

TBA. A software based trancode is optimal for non-live workflows, as they always give the best quality-per-bit. We also have signals from MistServer customers & projects that these kinds of jobs could open the Livepeer network for a new segment of customers, as they currently run expensive clusters and take a long time to do these kinds of tasks. This feature should focus on allowing node operators to offer different parameters of transcoding tasks (price, quality, speed), so that a Gateway could request long-running, non time-critical transcoding tasks.

TBA. Current issues are mostly aimed at O/T operator. We need to identify wants & needs from current Gateway node operators to put on the roadmap or bounty board.

TBA. If a transcoder shuts down, it drops all streams. It would be benificial if on O/T shutdown, Gateway nodes would first start a new transcode process and after that shut down the transcode session for a graceful handover.

TBA. Currently, full segments are sent between nodes using regular HTTP calls. We should investigate the benefits of using QUIC for this and if there are significant gains to be made, scope out a solution.

TBA. Gateway nodes should be capable in dealing with media. We should identify additional ingest or delivery options to increase compatiblity.

TBA. Livepeer currently uses a segmented workflow, where video streams are split into discrete segments. Each segment must be fully received, transcoded, and sent back before it can be sent to viewers. We aim to transition to a streaming workflow, where video frames or small chunks are processed continuously in real-time. This approach significantly reduces latency, providing near-instant feedback and smoother performance for interactive and live content. It’s a delicate, large-scope change that will require a more thought out plan.

TBA. There are a lot of codecs, transport protocols, etc. Newcomers to the media industry need to understand the inherent pros/cons of each of them and get a gentle introduction into the complex world of the media industry.

TBA. Media applications sending in valid RTMP media data should always be accepted by the Gateway node. Currently there seem to be issues with RTMP ingest, like having strict requirements on the keyframe interval. Issues need to be identified and fixed.

TBA. Especially for remote O/T setups (different code paths from a combined setup), we've seen cases where an issue during transcoding is classified as 'fatal', causing the entire Transcode process to restart. Dropping all streams is bad - we should find out the root cause of these restarts and try to prevent them.