Quality and Video Encoding

The vids I make are average in terms of picture quality as I prefer the lower file size over quality (512kb/s to 768kb/s). If there are any special moments, then I can increase the bitrate to improve clarity.
My PC has a 1TB hdd. Thats quite a lot for a casual pc user. For Vid Makers/uploaders, that's nothing. I can go though almost 100GB a day in raw data from recordings from Sky, the various streams and general downloads. Average is about 60GB a day. I just dont have the space or time to make vids of higher quality, nor do I have the desire to do it on a regular basis.

As a casual viewer of the babechannels, most the stints the various girls do across the channels can get a bit repetitive. For me, I dont see the point in consuming my pc's cpu usage or memory in creating a HQ vid. The vids are not unique as the same babe will most likely be on the next night doing the same thing.

I am now very selective on what I upload. As a vid maker, I only upload vids that I would want to download myself. Although I use a low bitrate, the vids I upload are still pretty good as can be seen in the caps I post. There is some degradation when upscaled, but not enough to warrent doubling the file size. I have a 40in Full 1080p HDTV as my monitor to check created vids as well as a standard multi monitor for the multi tasking vid recording process. Thr general picture quality of the channels is pretty poor. I improve the picture quality to a decent and acceptable level.

After I make a vid, within a week it is on my external hdd. Even for me its rare to watch old/older vids. I assume its similar for the people that download vids.

Plus, although ISP's say downloads/uploads are unlimited when you have high speed Broadband up to 30Mb. There are fair usage policies and I have a few letters from my ISP to say I use too much for both. Even if I wanted to, I cant anyway at the risk of losing my Internet connection.

I have evolved as a vid maker in my time on the forum. I knew nothing when I started and uploaded some crap in my time that still makes me cringe. I'm content and comfortable in my current vid making style and in the amount I upload. I dont spend as much time making vids or even watching the channels anymore. There is more to life then spending every spare minute editing and making vids.

As for using the graphics card to encode vids, unless you use over 1500kb/s, the quality is poor. (Although it's very fast) I have an Nvidia cuda graphics card.

Anna Elite Vids. ::: .EliteTV Vids. ::: .Kylie Tyler Nights Vid. ::: .Last Danica vid. ::: .Lolly/Paige 2-4-1. ::: .Last Paige Vid. ::: .Fernanda NightShow 1 ,2 ,3 ,4 ::: .Elite-Fancy Footwork. ::: .Elite-Freaky Like Me. ::: .Elite-It's Like That. ::: . *Evelyn Highlights Vid 1,2* ::: .My video bam Vids.

Thanks Josh. As we've discussed offline the point is to have a life outside these channels, and it's not all there is to the day.
I go through similar sorts of bandwidth although I have one saving grace with my ISP - uploads are unlimited and they truly are. Downloads on the other hand are capped and they charge £1 per Gb you go over, so I never go over !

I'll need to go and review the Cuda stuff a bit more it would seem as it looks like things have moved on a bit since I last looked in it...

Perhaps I'm falling back into my bad habit/traits of a perfectionist. There's little point as this stuff isn't for posterity or anything - just for those that can't see or missed a show to get a glimpse of what was on.

I don't personally target a bitrate anymore given the advancements in the x264 encoder - I just use CRF and set this to 24 (as a reasonable number) and let it do it's magic with the other settings I've used to help refine the amount of data it has to offer the encoder. For some other stuff where quality is paramount (not stuff I post here), then I use CRF 18 for close to transparency with source.

For now, I'll finish up with a quick overview of the quality from 6 different sampled encodes on a short vid I did earlier for my own understanding of the Quality vs Speed debate. Suffice to say that unless I really case about quality to the nth degree then I should just be looking to use the column 1 settings from below as that is >60fps whereas the column 2 settings max out even with MT AviSynth at 22fps. All processing was done in the same session within StaxRip on my AMD Phenom II x6 machine.

This image was taken from 6 different types of encode (basically 2 different interlacers and 3 different post-processing options).
All other encode settings/parameters were identical.

Clearly posting a JPG doesn't give you the level of quality for a real review/appraisal, so attached is also the files in BMP format as direct screen outputs from MPC. screenshots.zip
Also here's the source file and the outputs from the encodes so you can see for yourself (or indeed use the same source for your own analysis etc). Source File and Encoded Videos.zip

Yadif DeInterlacer, No Post-Processing (1.52Mb - 1342 Kbps - encoded 237 frames, 36.64 fps, 1255.78 kb/s)	QTGMC (Preset=Faster), No Post-Processing (1.48Mb - 1307 Kbps - encoded 237 frames, 11.43 fps, 1221.48 kb/s)
Yadif DeInterLacer, Post Processing CPU=4 (1.43Mb - 1270 Kbps - encoded 237 frames, 36.77 fps, 1183.76 kb/s)	QTGMC (Preset=Faster), Post-Processing CPU=4 (1.40Mb - 1237 Kbps - encoded 237 frames, 11.14 fps, 1151.40 kb/s)
Yadif DeInterLacer, Post Processing CPU=6 (1.37Mb - 1216 Kbps - encoded 237 frames, 36.57 fps, 1130.46 kb/s)	QTGMC (Preset=Faster), Post-Processing CPU=6 (1.33Mb - 1179 Kbps - encoded 237 frames, 11.15 fps, 1092.79 kb/s)

My "normal" settings for quick encodes are in row 2 column 1
My "normal" settings for quality encodes (and time not being an issue as it's anything up to 5 times slower and usually at least 3 times slower) are in row 3 column 2
My recent posts (e.g. for RLC Tease Time and Sammie extracts) have been using the settings for row 3 column 2 as well

I'd say for vids quality is more important than size 2 me.


As for the caps rota it could work among the regular cappers but for people like who only cap when they have time it could be hard 2 stick 2. One thing i have been doing lately is capping the girls that usually dont get as many individual users that cap them which sometimes go uncapped when up against the 'popular' girls, but as I only cap the elite girls i can predict that a bit easier

It just made me wonder if file sizes have got something to do with the quality of the video? Say, if i zip an excel file and open it, it remains unchanged. how about video files?

(11-02-2011 22:19 )ron77hays Wrote:  It just made me wonder if file sizes have got something to do with the quality of the video? Say, if i zip an excel file and open it, it remains unchanged. how about video files?

Video is already compressed due to the nature of encoding.
So whilst it's technically possible to ZIP or RAR a video file (MKV or MP4 for example), you will get <1% compression on this so no point (in fact more often than not it's the exact same size after putting in archive header and any CRC information). The source files (MPG) usually compress a little better for some reason at around 4%.

Well Dirk is absolutely correct in what he said about complicated set design, and intricate patterned outfits.

I've had some 2 hour videos fit quite comfortably within 400Mb. And yet I've had some 1 hour videos take up twice that, just because the babes been wearing a patterned outfit, and it's been a high detailed set.

Elite's mural background sets are a good example. Plus for some reason, most of the Elite babes seem to like fishnet stockings or bodysuits.

As for caps. I never do live screen captures. I always take mine off my stream recording, or off my finished video. So my caps only get posted, as and when I get around to editing my original recording.

Plus, I tend to run at least a month behind on most of my posts. And I still have a huge amount of stuff, still to edit up as I tended to record anything up to 3 webstreams at the same time, plus my satellite recording. So there is no way I could get all that edited up, uploaded and posted as quick as most can.

So where I'm concerned, it's most likely going to be a back recording I post, rather than yesterdays recording.

I'm probably more quality oriented than most when it comes to encoding.. and downloading what others post for that matter. I would never disrespect a contributor by bringing it up.. but it does disappoint me when I download a clip that isn't de-interlaced, has pixelation because the bit rate was way too low or the aspect ratio is out. I feel like I should PM the person and help them out but I don't want to come across as a dick.

Everyone is different when it comes to quality though. Some will spend weeks fiddling with a new TV to reach what they think is perfection, others won't particularly care if everyone on screen looks like an Oompa Loompa.

When I'm encoding a clip quality takes priority. I don't want anything lost (perceptively anyway) compared to the original. So a reasonably high quantizer, other quality settings on the high side and no reduction in resolution. When I started encoding it was just to turn my own recordings into finished vids, posting here started a lot later. Usually I stay with quality based settings, but sometimes the bit rate goes very high for reasons that have been mentioned. Complex sets and outfits etc, in which case I'll do a second encode with lower settings for upload, if its something I think people will want.

About de-interlacing, save your sanity and stay with yadif. I spent days figuring out how to use a more advanced and much slower de-interlacing filter with avisynth and didn't see enough benefit to warrant the effort. Babe channels aren't broadcast in high enough quality for it to be noticeable I think.

Enough rambling though.. to the actual questions asked in OP.

1. Do people care about the file size ?

Within reason, no.

2. Should vids be created at set sizes and take into account define profiles ?

Fixed sizes don't work too well. You either end up with a bad quality encode if a complex clip isn't given enough space to work with, or in the case of a 'simple' clip you might give it more space than it needs.

3. For those creating content, do you care if its MKV or MP4 or anything else for that matter ?

I don't think the format matters for people playing back clips on their PC. Bluray players that play mpeg4 might need one or the other but I don't know. I just use .mkv because everyone else does and I'm assuming there's a reason.

4. To keep quality aligned to file size, I find 30 minutes = 200Mb of file size. Is that the same for other encoders ?

See 2. But 95% of the time, at 720x400 that would work well actually. I don't see a reason not to play it safe and use a quality level though.

5. Is there any benefit to some of the encoders getting together to define a "standard" so that everyone will know what to expect when downloading something ?

I'd say its more important just to share knowledge and tips with others so they can do good encodes.

6. Should encoders share software used, techniques, settings etc ?

Sure.. why not.

7. Instead of lots of people capping the same channels, should common encoders create a schedule of who will record what channels and post what content ?

That's a good idea actually. I don't watch or record as much these days but could certainly do one of the freeview channels if it would help.

On the whole deinterlace piece this has taken me a considerable amount of time to investigate properly as whilst you'd think this was easy and just pick the "best bits" from both frames and weave them together. However such an approach tends to leave a nasty image behind.
I agree Yadif is by far the best and actually simplest to use when working with medium quality aired content such as the babe shows. And indeed up until about a month ago that was all I used.
But with the recent way in which sets are being put together on the shows this was showing signs of not coping well, hence my investigation and ultimate use of two different solutions dependant on whether I want to wait the time for the encode vs getting the stuff done. This means two different products as I've never been able to find the same settings for one library based deinterlacer as I use in AviDemux to the ones I could use in AviSynth (which is weird as it should be dead easy).
In AviDemux I now use Decomb Telecide with No Strategy which gives me close to Yadif speed (only around 5% slower) but much improved (when looking at a single frame I'll grant you so not as obvious when watching content as video) output and same size/kbps and a sharper image.
In StaxRip I have tried to use the equivalent deinterlacer and I don't get the same quality - possibly my skill at trying to work with AviSynth scripts and not a fault of the plugin. Therefore here I use Yadif for quick stuff, and then QTGMC v3.01 on Faster setting for quality (and not caring it takes 3 times longer).
The main advantage of QTGMC is that is does indeed process fully both frames of the deinterlaced picture and it does determine through some clever maths which bits are "probably" best. You then (for PAL) do a SelectEven() and you end up with the best of both fields into that Even only field for the actual encode. This allows x264 to process a cleaner frame and hence quality goes up and interesting kbps goes down (as the cleaner image needs less processing to encode).

For encodes you'll notice that I do indeed resize my video and I ensure I'm as close to 0% variance to the original ratio after cropping (can't always get to 0 but try to stay within 1%). I know that any resize affects quality as your effectively throwing away good video data in the process. However I did trial here some time back no resize at all and to then use DAR in the MKV Muxing processing to set this correctly - but people didn't relate well to that type of encoding so I reverted to resizing it.

For me quality comes first and file size is not really relevant. I use CRF encoding and then limit the quantizer to fixed lower and upper maximums as well as defining the maximum qp change to 1. This ensures smoothness of encode quality and perception of the quality of the output goes up. With high movement or high detail background (or nets on the girls) far too many kbps are wasted on stuff you can't really see. On other stuff (like the other nights Danica show) you can get a very high quality output at an average of only 680kbps as the movement is fluid with sensible backgrounds etc. Perhaps the most annoying bits are the scrolling text - in tests I've been able to evidence this accounts for between 8 and 12% of the overall encode size and allocation of bits - just for text you can't read anyway (I've done the tests using the same source file, just with AviSynth scripts to use overlay to put a black box over the text scroll areas - this isn't something I post as this would be against the aim of the content owners who grant us the freedom on this forum to post any video content as they do own the copyright etc).

I'd really like it if the producers of the shows understood that to give out a high quality product they themselves have to think about their own bandwidth through the $ky transponder and with VBR allocation peaks of only circa 3.8Mb (absolute max, most trundle along around the 1.5 to 2.2Mb range) they need to consider this as well. Given the weird DVB-S resolutions used (mostly 544x576 with a DAR force to 16:9 to force your TV to "fix" this out to the correct size) you've not got a lot of pixels to work with.

Okay - I think I've waffled enough for today. Suffice to say that people have differing views as to what to do and why and they're all perfectly valid as there are different requirements and aims from the people doing the work.

I use AviDemux with Decomb Telecide for quick encodes or where I've a need to perform complex editing (as it's very good at this).
I use StaxRip with complex AviSynth plugins and options where quality is required or where I don't mind waiting a bit for the encode to complete.
I only ever work in MKV files as they're just the encapsulation of the underlying video/audio in the MKV container and have good options for other things as well as good general tools (specifically MKVtoolnix).
I only ever work in CRF encodes now as 2-pass doesn't seem to make any difference and just takes longer.

I'm happy to share my StaxRip profiles and templates for those that might want them, or indeed my AviDemux custom jobs. These provide all the settings I've taken so long to cultivate into the videos I produce today from the babeshows.

Might give QTGMC a try then. A quick google shows it's based off TempGaussMC which is what I played with a few months back.

The odd 544 x 576 res is used by all the Freeview channels too. Obviously the smallest encodes come from leaving it at that and letting the playback software resize it. The problem is most software uses a quick and dirty scaling method that doesn't seem to like such resolutions much, so the encodes would look poor for most of the people who download the videos. So I do a lanczos3 resize to 1024 x 576, or 720 x 400 if needed.

Having done more test encodes with and without some OSG/scrolling text I think the overhead this adds to the kbps to keep quality is significant enough to do something about it (ie remove it). It can be anything between 8 and 25% kbps reduction for exactly the same video with exactly the same encode settings. That's enough to make a difference so I'll do a few real posts with these done and see if people like/hate/don't care...

Oh and yes I appreciate in one of my earlier posts in this thread I stated I wouldn't do this. But the file size difference speaks for itself. A sample 1 hour encode of Laura Live is only 68% the size of the encode if I leave the scrolling text in (accepting the ratio won't be constant with different source videos as there are other factors to consider such as babe movement, background etc). That's a massive amount of difference and I'm sure will help when people want to download stuff but don't 'cos the files are too large.


A couple of example scripts:-

For Laura Jones (on TopShelfTV)

For night time shows on Club Paradiso