How to read results of the Compressibility check?

[Bahamuth]

Greetings everyone,

I'm afraid I haven't found an answer with the search function, so I wanted to ask something.

Could one use the result of the compressibility check to adjust the CRF or Bitrate quality of the movie to match the original inputs compressibility? I know it's mainly used if you want to achieve a specific end file size with 2pass encoding, but I'm still curious about the topic.

I made a few tests (I even raised the percentage for the compressibility check to 20% of the movie):
Duck Tales DVD: 0,56
Mary Poppins BluRay: 0,46
John Wick 2 BluRay: 0,11
Pokémon 21 BluRay: 0,13

I know that Duck Tales and Mary Poppins are somehow rather noisy movies, thus I think the compressibility is "bad" because of that. (Some screenshots from Mary Poppins to be found at https://silentstrider.net/mp.jpg saved with highest JPG quality not to give false results from jpeg artifacts). The other movies are of course one anime, but also one action movie, but both "made" for BluRay.

So.... can I use a "lower" quality setting (CRF 23 or higher) for the movies with lower compressibility results and still get a good result?

[FranceBB]

I wouldn't personally go above CRF22, especially on grainy sources.
As a broad view, I consider CRF 15 as optimal quality, CRF 18 as good quality, CRF 22 as acceptable quality; anything above CRF 22 is not so good and when I made a survey, even people with untrained eyes spotted artifacts at CRF 25.
As to Duck Tales and Mary Poppins, you can use --tune grain and you may also wanna play with --aq-strength as it can change the perceived quality, but since "perceived quality" is relative to the user, you have to choose what it is looking best for you.
In a nutshell, an high --aq-strength value retains more gradients (i.e avoids banding) but it's prone to produce ringing, while a low --aq-strength value preserves edges but it's prone to produce banding.
I would suggest you --aq-strength 0.4 for grainy sources and --aq-strength 0.8 for anime.

A few additional questions: What's your target result? Do you care about TV compatibility? What preset are you actually using? Can you share your full command line?

[Bahamuth]

Mary Poppins is a very strange movie, because I remember encoding it with something like CRF19 and the result was bigger than the original.

So CRF seems rather unstable for me, and with having only an i7 3770, I wouldn't want to encode just one movie over multiple days again and again to find an "acceptable" size, even if it's only for a few movies.

Anime movies in general always seem to end in 1/10 of the original size, but real movies (yet without the preset settings like for grainy movies) seem to differ greatly in the resulting size.

So I guess there is no "acceptable" bitrate for something with 0,11 or even 0,46 which I could set as CBR? In another thread I have read that Life of Pi is also rather high with the compressibility check and I doubt that's because of grain in the movie.

[Asmodian]

Some things compress well, some do not. There really is no general bitrate for acceptable quality, as you see the range of bitrates required for similar quality is very high. This is why CRF is helpful.

Water is very hard to compress too, I imaging this is the cause for Life of Pi.

[Bahamuth]

Would the movie compress better if I add a denoise filter to it? Like making textures more compressible again?

[ChaosKing]

Would the movie compress better if I add a denoise filter to it? Like making textures more compressible again?

yes, you could try some light denoising. Try Smdegrain().

[Bahamuth]

Summarized that means:
From the compressibility value I cannot solely tell, if I should use a higher or lower CRF. Neither does it mean, that a high compressibility value is a more complex video, it just can be noisy, because it's just an old movie (or in case of Duck Tales because of mpeg2 compression I believe. I made images here: https://silentstrider.net/ducktales.jpg). Adding a denoising filter might help, but I should not set the filter too high (I know Disney did that on many BluRay movies).

I could make a compressibility check and upon a high value, I could try to check, why, e.g. noisy or some other reason (water e.g.), but that might take more time and shouldn't be the standard. Current Anime BluRays should never have that problem, Anime DVDs might have, current Live action movies on BluRay should usually work with a "decent" file size in the end (e.g. not near original file size), Live action DVDs might be tested. Luckily it's faster to make a compressibility check von DVD movies because of the image and filesize.