Can't even get near Netflix Silverlight Quality

[pzo]

My quest is pretty simple: Conversion of ordinary digital 4:3 AVI camcorder output to a smaller file size w/o losing (too) much quality. Two file sizes, actually. One for sharing/streaming, and one for high quality viewing.

I've spent probably two days experimenting with, mostly, Any Video Converter and Handbrake, all kinds of settings including MPEG-2, x264, and FLV. I try variations in frame rate, deinterlacing, "Quality" in Handbrake, frame size, you name it. If a setting can be changed, I've tried it. I use a simple 8 minute test video, look at the resulting file size, then look at the video, and the MB/sec result.

I've used Gspot and Media Tab to analyze downloaded videos for analysis and hints what works, quality v. file size. I have no need or interest in anything grander than a computer monitor and 720x480. No HD, no nothing fancier.

The best I can come up with is a set of settings on Handbrake that give me "acceptable for sharing, minimal file size" results at 2.7mb/sec. The best I can come up with for "virtually no degadation" is also via Handbrake at 6.4mb/sec. Generally, every video download I've checked, including some obviously very professionally, expertly done ones, fall into that range, too.

And then I look at streaming Netflix. My DSL is the lowest speed permitted, 1.5mb/sec, and the image is as good as would ever hope for or ever need. (Again, I don't want, need, or care about HD, blah, blah.) Per Wikipedia, this is the lowest quality setting that NF will stream. It's Silverlight H.264, according to the same article.

So how does Netflix/Microsoft/Silverlight encode at 1.5mb/sec in a quality that I can't get below 6.7mb/sec or more? I mean, that's a 1:4 ratio, not just some little nuance or quibble.

Your thoughts, and hopefully, help, please!

[poisondeathray]

My quest is pretty simple: Conversion of ordinary digital 4:3 AVI camcorder output to a smaller file size w/o losing (too) much quality. Two file sizes, actually. One for sharing/streaming, and one for high quality viewing.

I've spent probably two days experimenting with, mostly, Any Video Converter and Handbrake, all kinds of settings including MPEG-2, x264, and FLV. I try variations in frame rate, deinterlacing, "Quality" in Handbrake, frame size, you name it. If a setting can be changed, I've tried it. I use a simple 8 minute test video, look at the resulting file size, then look at the video, and the MB/sec result.

I've used Gspot and Media Tab to analyze downloaded videos for analysis and hints what works, quality v. file size. I have no need or interest in anything grander than a computer monitor and 720x480. No HD, no nothing fancier.

The best I can come up with is a set of settings on Handbrake that give me "acceptable for sharing, minimal file size" results at 2.7mb/sec. The best I can come up with for "virtually no degadation" is also via Handbrake at 6.4mb/sec. Generally, every video download I've checked, including some obviously very professionally, expertly done ones, fall into that range, too.

And then I look at streaming Netflix. My DSL is the lowest speed permitted, 1.5mb/sec, and the image is as good as would ever hope for or ever need. (Again, I don't want, need, or care about HD, blah, blah.) Per Wikipedia, this is the lowest quality setting that NF will stream. It's Silverlight H.264, according to the same article.

So how does Netflix/Microsoft/Silverlight encode at 1.5mb/sec in a quality that I can't get below 6.7mb/sec or more? I mean, that's a 1:4 ratio, not just some little nuance or quibble.

Your thoughts, and hopefully, help, please!

Simply put, it's not possible

The short version is "garbage in = garbage out" . You're comparing Apples and Oranges.

It has very little to do with Netflix encoding, and their quality and compression is actually fairly poor. On a technical level, the actual encoding and compression can be better with x264. So the differences are in source quality content and acquisition format

Nobody is using DV-AVI to make a decent shoot for 10-15 years, even a low budget production. Even the crappiest HD cameras can produce nicer, cleaner picture than DV camcorder. Oversampling has huge benefits - a noisy, low detail image at UHD or HD can look superior when denoised and downscaled. Typical DV camcorder signal to noise ratio is very low - both in terms of crappy sensors and recording compression. Your starting with a huge handicap with that alone

To give you a simple explanation, modern compression works by storing differences between frames. The larger the differences, the more difficult to compress (ie. larger file sizes for a certain level of "quality"). The content and shooting technique will be much different. A professionally shot, stabilized, professionally lit studio recording will be much easier to compress than some dark, shaky, handheld home video.

Most theatrical productions will be film or digital film equivalent shot at "24p" . Consumer video will be interlaced 29.97 (or bob deinterlaced to 59.94p) , even if you ignore all the mentioned major differences, you will require more bitrate for a certain level of quality because of more samples / second - but the motion will be much smoother and more forgiving. Also shallow depth of field will improve compression considerably in the theatrical, professionally shot case.

If it was poorly shot, you can improve compression by stabilizing, denoising, using better deinterlacer. Do you recall in your other thread, the types of things that were mentioned ? You always need to do extra processing for DV-AVI. The better the deinterlacing, the better the denoising, the more optimized and easier the compress. Handbrake doesn't have very good deinterlacer or denoising. You have to venture into avisynth land with QTGMC for better deinterlacing and access to more and better denoisers (learning curve) , or other programs like neat video for denoising - and even that won't make a huge change because there are just to many differences between the content and acquisition (apples and oranges). I don't even have to look at what you have and I can tell you that is true.

You will never obtain as clean or high quality or easy to compress with what you have, there are just too many differences. You can improve the compression ratio slightly by some of the things mentioned, higher compression settings, but the main difference is the source and acqusition characteristics