Info wanted: General Encoding Question from Projects - Scripters opinions wanted.

I’m searching for a way to encode a file for best possible constant quality, based on a specific size.

I was wondering if there was a way to tell an encoder to CRF up to the credits (specifiying a specific frame), at which point the credits are encoded using two pass to achieve an overall average bitrate, acknowledging the size of the CRF encoded portion as well.

You would set a minimum bitrate so that the credits weren’t totally botched. The encoder would look at the CRF portion, then see if the credits could be encoded at the minimum bitrate to reach the desired size. If not, the encoder would repeat the CRF encode at one value lower, and repeat the process. Otherwise, the two pass would be completed on the credits to allow for potentially better quality

All of you scripters… has this been done, or is there any way that you could look into a way to do this for mpeg2 and h.264?

Thank you for that, it helps a lot. I see your point on the credits - while it may seem useless to some, perhaps a simple script that checks the final size of a CRF encode, and discards it if over a certain size, starting over to encode again with a CRF value one lower (a smart script might even be able to judge how much larger the encode is than the desired size, and choose between 1, 2, or 3 lower depending on the overage).

Oddly, I see the potential in zones evening out quality in projects using different sources. Maybe CRF 16 for Upscaled HD footage, while bluray footage gets CRF 20+ to degrade the quality of the film through encoding the same way SD footage was originally degraded through encoding. I'd imagine this would work far better than any unsharp tool.

No, I think you're better off using a consistent CRF or a consistent 2-pass bitrate setting. There are other ways to degrade picture quality without using different bitrates - for instance you could resize the scenes surrounding your inserted SD footage to make the transition more smooth, if desired (downscale to a resolution in-between and then upscale it again to degrade the quality).

The credits take up a negligible amount of space compared to the rest of the film anyway. So I agree with junh1024 on that one. For instance, in my encode for Jurassic Park the video file came to 3542.2 MB, yet the credits are only 147.1 MB of that. So by compressing that area of the film further I could only really save an extra 50 or 60 MB and that's being generous. Out of 3.45 GB, 50 MB is nothing.

However, he's wrong about 2-pass. CRF and 2-Pass are equivalent. The only difference is that CRF maintains consistent quality and achieves the desired level of quality, whereas 2-Pass maintains consistent quality and achieves the desired average bitrate. If you have a 2-Pass file and a CRF file that come out at the same size then they will be exactly the same quality. Which you can test yourself easily - set CRF quality to something low like 26 or 28 and encode, then take the average bitrate from the file you just encoded and encode again using 2-pass. The two encoded files will be exactly the same quality - they just came to it using two different routes.

Here's another tip... with 2-Pass you can't know the quality in advance. So my suggestion is you set the CRF to what you want and encode the file. Let's say you want the CRF at 18 or perhaps 19. Encode it and see what size it comes to. If it's well outside the size you're wanting then you should find a way to make the file more compressible instead of lowering the encoding quality to force it to fit the desired size. You could resize from 1080p to 720p for instance. De-grain, De-noise, etc. Once you get it to encode around the right size at the right quality then use 2-pass to get the exact size that you need.

My experience is that x264's "2-pass" mode is broken and should not be used. If it worked like xvid's 2-pass that would be another story.

Well xvid didn't have crf (constant rate factor), it only had cq (constant quantizer) which is not the same. x264's 2-Pass works exactly as it should, if it didn't then CRF wouldn't work either since CRF is based off the same principal as 2-Pass - that is it keeps constant quality. CQ does not... it aims to keep the quantizers to a constant level, even when a scene doesn't need it, therefore 2-Pass will give you a better quality output at the same size as CQ-based encode.

Is there a way to tell MeGUI to only encode between certain frames?

e.g. encode between frame 1000 and 2000

I say because from experience 2-pass doesn't actually appear to *do* anything with the data from the first pass - the running bitrate average is like that of a 1-pass encode.

I don't go there. But I've talked to other more experienced anime encoders and they almost invariably say "CRF or bust", unless you're making BDMVs.

In theory, at least.  In practice, that's another story - I can't think of any instance where analyzing the source file should suggest the proper way to handle any file as almost CBR, which is what x264 almost universally did when I used its 2pass mode.

With a file I encoded not long ago I compared the bitrates and it's only slightly "more cbr" when encoded with 2-pass. That could be due to the faster first pass settings, I certainly didn't use --slow-firstpass after all. But overall does it look like CBR? No it doesn't, it looks like a proper VBR.

Because I needed it to fit on a DVD-R.