The current docker image is quite heavy ~1.5GB and it makes sense as it contains utils needed for the hardware transcoding.
However, most of the userbase doesn't have a Nvidia GPU on their hardware since it mostly runs on mini pc/NAS that have either no encoder or just small intel QSV. And especially with the current nand economic context, ssd space in homelabs must be used carefully.
What I would suggest is to keep only ffmpeg for cpu encoding in the base "latest" image, and instead provide latest-nvidia (or latest-hwtranscode) for those who would like to have hardware transcoding.
Or if you don't want any breaking change, have a tag like latest-light for those who don't want any hardware transcoding
The current docker image is quite heavy ~1.5GB and it makes sense as it contains utils needed for the hardware transcoding.
However, most of the userbase doesn't have a Nvidia GPU on their hardware since it mostly runs on mini pc/NAS that have either no encoder or just small intel QSV. And especially with the current nand economic context, ssd space in homelabs must be used carefully.
What I would suggest is to keep only ffmpeg for cpu encoding in the base "latest" image, and instead provide latest-nvidia (or latest-hwtranscode) for those who would like to have hardware transcoding.
Or if you don't want any breaking change, have a tag like latest-light for those who don't want any hardware transcoding