They recently added NVenc support in addition to the previous CUDA accelerated transcription runner. I didn’t have time to test it again yet, but my NVenc setup inside the main Peertube container works great since the 8.x upgrade so this should be also useful.