Nit: here it might be useful to explicitly say that the batch methods are faster than the single-frame decoding methods - e.g.get_frames_at() is faster than callingget_frame_at() multiple times.

Above:

This is a good description. I think we can be more nuanced about when to recommend approximate, e.g. we should try to clearly articulate the last 3 bullet points which are currently slightly overlapping and contradictory (we now know that approximate won't always be "a net win").

That's on me: I need to first have a clear understanding of why approximate mode is sometimes slower, and I'll need to update the approximate mode tutorial with more detailed recommendations.

I won't be able to do that in the next few days, so to unblock yourself I think you can just remove the claims about approximate being strictly superior ( bullet points 2 and 3), and the more generic reco could be something like

If the video is long and you're only decoding a small amount of frames, approximate mode should be faster.

It's not super actionable for users but I hope the dedicated tutorial I'll edit will be more precise.

I think it's good to have those bullet points here. They overlap with what is already in the CUDA decoding tutorial, and I think we'll want to remove them from there and have them here instead.

Eventually we'll also want to update the CUDA tutorial to explain to users how to check whether they're falling back to the CPU.

Mainly here in this tutorial, I think we should insist on one thing (as the main point): users should be using the Beta interface with

withset_cuda_backend("beta"):dec=VideoDecoder("file.mp4",device="cuda")

Maybe it would be more clear to group the similar functions here?
For example, we could add two small headers to group index based vs timestamp based retrieval:

For index based frame retrieval:

• get_frames_at
• get_frames_in_range

For timestamp based frame retrieval:
...

super nit: this is a personal writing style preference, but within a section lets consistently use either active or passive voice. For example, we could remove "you" from the first bullet point, and instead use the passive voice: "bit-exact results are needed"

Let's consistently use active voice. :)