🗳️ Requests

TBA. Currently, from the perspective of a Transcoder, transcode sessions get distributed round robin with a static session limit. This causes the encoder or decoder to get overloaded in some cases. An initial improvement could be benchmarking encode and decode pixel throughput and limit based on that. Even better: watch encoder and decoder usage directly, but this is Intel/Nvidia/Netint specific, so could be added in phases.

TBA. This feature was almost completed. There are branches which might only need a rebase to the latest go-livepeer version.

TBA. This would allow relatively modern Intel CPUs and Intel GPUs to be used for transcoding! The feature description should include benchmarks. Target devices with an Intel Arc (embedded) GPU and last-gen CPUs (Intel CPU with UHD Graphics 750/770?)

TBA. This feature has already seen significant work from as-astra. Might already be at a code-review stage? We should make it a bounty and help them get it merged!