Why HDR cannot use 8-bit
Bit depth is not about how many colours a scene contains, it is about how finely the range is divided. BT.709 spreads 256 levels across a 100-nit range. PQ spreads its code values across up to 10,000 nits. Attempting that at 8-bit puts visible steps between adjacent levels, which is banding, and it appears exactly where it is most obvious: in smooth gradients like skies and studio backdrops.
10-bit gives four times as many levels to spend, and the PQ curve is shaped so those levels land where the eye can actually tell them apart.
What converting an 8-bit source gains
Converting 8-bit footage produces a genuine 10-bit file, and that is worth something: the conversion itself does not add new quantization steps, and the extra levels give the remapped highlights somewhere to sit without introducing new banding.
What it does not do is recover detail the 8-bit original never recorded. If the source already has visible banding in a sky, that banding was baked in at capture and it will still be there. See converting 8-bit video to HDR for the honest version.
In a file
The pixel format carries it:
ffprobe -v error -select_streams v:0 \
-show_entries stream=pix_fmt -of default=nw=1 input.mp4
yuv420p10le is 10-bit. yuv420p is 8-bit. The 10le suffix is the tell.
Questions people ask
Do I need a 10-bit monitor to benefit?
For HDR playback, effectively yes, and most HDR-capable displays are 10-bit or 8-bit with dithering that approximates it. For editing, a 10-bit source still helps on an 8-bit monitor because the extra levels survive grading and re-encoding even when you cannot see them directly.
Does 10-bit make files much bigger?
Somewhat, typically in the region of 20 to 30 percent for the same perceived quality, though it varies with content and encoder. It is worth watching: an HDR rendition that came out smaller than its SDR counterpart has usually lost the highlight detail the conversion added.