h264 MP4 Encodes

[glenpinn]

Hi there, i hope i can post this request up here as i would like to ask someone to download some h264 MP4 encode files that i created using Handbrake 0.9.8 & Avidemux

i have 5 files (various sizes) uploaded into my Mediafire account if someone would like to take a look at them.

What i am after is for someone to check the sample source file, and then take a look at the other 4 files encoded at RF20 & RF25 and offer me an opinion about the output quality of the individual files.

The source file is a 40sec clip filmed outside my home using my new Panasonic HDC-SDC750 camcorder, which was filmed in 1080/50p mode (AVCHD m2ts @ approx 26,000br)

Settings for Handbrake (cpu encoding) Avidemux (cpu encode)
format = h264 MP4, constant framerate, framerate same as source, constant quality using RF20 & RF25 settings, with aac stereo audio 48,000Hz @ 256br

my encoding PC runs an intel IB 3770 CPU (HD4000 graphics) + 16gb ram + 120gb intel 520 SSD for windows/apps

- Handbrake took 40 seconds to encode the RF20 file using 60 to 70% of the cpu.
- Avidemux took 60 seconds to encode the RF20 file using 99% of the cpu.

it appears that Avidemux and Handbrake both have similar bitrates at RF20 and RF25 quality.

these are the file properties screenshots taken from the source file and the 3 encodes at RF20 (disregard Mediaconverter) i have lost all faith in this program.



this is the GSpot screenshot for the Source file and 3 mp4 files, and in these shots, Avidemux shows up as being 100 frames/sec where in the file properties screenshot it shows as 50 frames/sec ??? so maybe its time to give GSpot the flick, and give MediaConverter the flick as well, or change its RF quality setting to around 26 or 27 to match handbrakes RF20 bitrate setting. (again disregard Mediaconverter)



Download link for sample files if anyone wants them (apologies for the size)

1 Source File = 40sec @ 119mb (26,152br - 50 frames/sec)
http://www.mediafire.com/?qu343qqmoh5ddmf

2 Handbrake @ RF20 = 63mb (12,741br - 50 frames/sec)
http://www.mediafire.com/?8go4h9c3anrsyi8

3 Avidemux @ RF20 = 49mb (11,106br - 50 frames/sec)
http://www.mediafire.com/?73eter3k9933azx

4 Handbrake @ RF25 = 26mb (5,133br - 50 frames/sec)
http://www.mediafire.com/?5og3pv2gzs60y7x

5 Avidemux @ RF25 = 24mb (5,406br - 50 frames/sec)
http://www.mediafire.com/?5qy4vy3ynhsjrvu

[poisondeathray]

maybe quicksync is limited to High@L4.1 ? It's capped to 1080p30. 1080p50 requires High@ L4.2

looks like the avidemux/ffmpeg transport stream bug - double FPS = 100 . It has problems with decoding transport streams

if gspot is accurate, none of the framecounts match

[glenpinn]

maybe quicksync is limited to High@L4.1 ? It's capped to 1080p30. 1080p50 requires High@ L4.2

looks like the avidemux/ffmpeg transport stream bug - double FPS = 100 . It has problems with decoding transport streams

if gspot is accurate, none of the framecounts match

hi, ok is the Avidemux "bug" a real concern showing 100 frames/sec or is that just a glitch given the file properties shows it as 50 frames/sec and GSpot shows it as 100

i will look into quick sync a bit more, if what you said about it being limited to High@L4.1 then there is no point in using it.

Also, i would be interested to know why MediaConverter has such a dramatically higher bitrate level at at the same constant quality RF settings in Handbrake and Avidemux, maybe its another glitch or side affect of using quick sync ???

any thoughts on the source file quality from the camera, or any of the encoded files ???

[poisondeathray]

hi, ok is the Avidemux "bug" a real concern showing 100 frames/sec or is that just a glitch given the file properties shows it as 50 frames/sec and GSpot shows it as 100

It's a real bug and a long standing issue with some transport streams and avidemux. Look at the file information when the video is loaded in avidemux. It "thinks" the source is 100fps, so anything it does afterwards will export at 100fps

i will look into quick sync a bit more, if what you said about it being limited to High@L4.1 then there is no point in using it.

It was just a hypothesis that might explain the 30fps . See if you can adjust the level and profile to high@4.2 somewhere in the program

IvyBridge QS is going to be very fast and expected to be faster than any software h264 encoder (even when adjusted for very fast settings), but when you encode ~70% more frames, the software encoder is going to be at a huge disadvantage if you are comparing speeds.

Also, i would be interested to know why MediaConverter has such a dramatically higher bitrate level at at the same constant quality RF settings in Handbrake and Avidemux, maybe its another glitch or side affect of using quick sync ???

RF is meaningless between different programs . Even between handbrake and avidemux which use the same encoder , x264, they use different base settings so CRF value is not comparable

any thoughts on the source file quality from the camera, or any of the encoded files ???

I've only downloaded the source file so far, I'll have a look at it tomorrow as it's getting late here

[glenpinn]

I've only downloaded the source file so far, I'll have a look at it tomorrow as it's getting late here

ok thx a lot, no rush, i understand its late evening/early morning where you guys are, here in sydney its 4.15pm friday arvo so things are just winding up here

i just like handbrake because it doesnt thrash my 3770 CPU at 100% like many other cpu based encoders do, so i can leave it on thru the night to encode, where i wont leave a program running that is thrashing my CPU at 100% like Avidemux does, and have it take a lot longer to ouput.

regarding quick sync, it is brutally fast over cpu based encoding (75% average) and it is just sick to see the output status bar moving so fast in mediaconverter over the slowness of handbrake

OH btw, have you personally got any preferences for specific encoding tools.

cheers

Mediaconverters h264 profile settings is shown here (Level is at 4.0 by default)

another issue with this program is that it only does Arcsoft AAC stereo audio @ 44100 or 48000, not that i have an issue with that, i like to use AAC stereo most of the time anyway.

EDIT: unfortunately no change to 30 frames at 4.2 or 5.0

[poisondeathray]

i just like handbrake because it doesnt thrash my 3770 CPU at 100% like many other cpu based encoders do, so i can leave it on thru the night to encode, where i wont leave a program running that is thrashing my CPU at 100% like Avidemux does, and have it take a lot longer to ouput.

If you change the settings to match (handbrake and avidemux), they will probably act the same way and take the same amount of time because they both use x264

It's funny , most people want 100% CPU usage to reduce slack/idle time and complain the other way

regarding quick sync, it is brutally fast over cpu based encoding (75% average) and it is just sick to see the output status bar moving so fast in mediaconverter over the slowness of handbrake

But in your test the CPU encode had ~1.7x more frames ... Try making the tests more comparable (and even then QS should be faster)

You can increase the speed of handbrake, avidemux (or any GUI for x264) by using faster settings. But quality and compression is the difference. And by all reports QS is still behind in those regards.

To do valid tests for "quality", they need to be at the same bitrate.

OH btw, have you personally got any preferences for specific encoding tools.

No, I use custom settings for different scenarios , content, and destinations . Sometimes compatibility is an issue, sometimes optimizing a certain characteristic like grain is the goal , you might use different settings for different content e.g. animation vs. grainy movie , etc....

Mediaconverters h264 profile settings is shown here (Level is at 4.0 by default)

EDIT: unfortunately no change to 30 frames at 4.2 or 5.0

What about changing to 4.2 and using something other than "same as source" for fps ? Can you manually enter "50"

Generally you want to use b-frames, they improve compression . Only if you use enormous amounts of bitrate will using no b-frames be more beneficial for quality . b-frames are lower quality (higher quantizer) than other types of frames, but at normal playback speed you don't notice the small differences so it improves temporal compression. At very low bitrates, using more b-frames definitely helps. Some devices and hardware players may have compatibility issues with certain settings . For example you can't use more than 3 consecutive b-frames for blu-ray

[glenpinn]

It's a real bug and a long standing issue with some transport streams and avidemux. Look at the file information when the video is loaded in avidemux. It "thinks" the source is 100fps, so anything it does afterwards will export at 100fps

sorry i meant to ask before, if GSpot detects it as 100 frames/sec why does it show in the file properties>details as 50 frames, and the footage from Avidemux still looks amazing even tho there is this apparent bug, so should it be advisable not to use Avidemux for encoding 1080/50p all together regardless of the apparent output and playback quality.

not that i personally would use it on big outputs because of its cpu usage being 100%, but for short clips that i cut from a source files i would use it for sure to output to MP4 because its already there.

[poisondeathray]

Can you post mediainfo (view=>text) on the avidemux encoded file, copy & paste the results back here

The avidemux is a DEcoding bug, not an ENcoding bug. ie. You shouln't pass judgement on x264 encoder - it can only export what it's given . If you used another compression format e.g. xvid , they would be 100fps as well

(The reason I'm not downloading the other files, is they are not comparable - the bitrates need to be more equivalent. It doesn't really demonstrate anything)

[glenpinn]

If you change the settings to match (handbrake and avidemux), they will probably act the same way and take the same amount of time because they both use x264

It's funny , most people want 100% CPU usage to reduce slack/idle time and complain the other way

if i can use handbrake at 60% to 70% cpu usage and get it faster than another program that uses the cpu harder, i will pick handbrake any time and there is nothing worse than using a pc while the cpu is belting along at 100%

But in your test the CPU encode had ~1.7x more frames ... Try making the tests more comparable (and even then QS should be faster)

i dont know how i can make the test between Handbrake and Mediaconverter more comparable except find a common bitrate, which i already tested and found that handbrake RF20 is 11,000 and Mediaconverter RF27 is about 11,000 as well, all other output settings are pretty much the same, and i dont know how i can get mediaconverter to get the 50 frames/sec output.

You can increase the speed of handbrake, avidemux (or any GUI for x264) by using faster settings. But quality and compression is the difference. And by all reports QS is still behind in those regards.

i think jagabo told me the other day in another thread to use faster x264 setting in handbrake because it generically uses a slower setting, but i posted up a screenshot of something, and he then conceded that maybe the x264 setting in handbrake may already be set to one of the "fast settings" already.

To do valid tests for "quality", they need to be at the same bitrate.

yeah now that i understand RF levels are not the same in each program, i can just test various RF settings for each program to find very similar bitrates between all programs.

What about changing to 4.2 and using something other than "same as source" for fps ? Can you manually enter "50"

see screenshot below, there is an option for "same as original" but no "50" option, so maybe MC is reading the source file wrong.

Generally you want to use b-frames, they improve compression . Only if you use enormous amounts of bitrate will using no b-frames be more beneficial for quality

i was told to use the default of 3 b-frames in handbrake regardless of bitrate ???

this is the media converter h264 frame rate settings

[poisondeathray]

if i can use handbrake at 60% to 70% cpu usage and get it faster than another program that uses the cpu harder, i will pick handbrake any time and there is nothing worse than using a pc while the cpu is belting along at 100%

These CPUs are meant to run at 100%. There is huge overhead and the chips run cool unless you increase the voltage significantly. You can overclock safely quite a bit even on stock cooler

i dont know how i can make the test between Handbrake and Mediaconverter more comparable except find a common bitrate, which i already tested and found that handbrake RF20 is 11,000 and Mediaconverter RF27 is about 11,000 as well, all other output settings are pretty much the same, and i dont know how i can get mediaconverter to get the 50 frames/sec output.

You need comparable framecount as well. If they are both "x" bitrate, but one video has more frames, you can thing of that as same amount of bitrate spread over more frames - i.e. each frame gets the equivalent of fewer bits

But that's what people do if they test CRF encoding - they do multiple encode iterations e.g. 21.1, 21.7, etc.... until the bitrates match

i think jagabo told me the other day in another thread to use faster x264 setting in handbrake because it generically uses a slower setting, but i posted up a screenshot of something, and he then conceded that maybe the x264 setting in handbrake may already be set to one of the "fast settings" already.

mediainfo view=>text will usually show the metadata and encoding settings used. It's easier to compare , say avidemux's vs handbrake vs. another GUI this way

see screenshot below, there is an option for "same as original" but no "50" option, so maybe MC is reading the source file wrong.

In some programs sometimes you can type something in there (you don't necessarily have to use the drop down presets)

Are there other GUI's you can use for QS that might not be limited in FPS ?

i was told to use the default of 3 b-frames in handbrake regardless of bitrate ???

That's safe for most purposes and generic use. Notice you had "0" for media converter. In a proper comparison , ideally you would use similar settings

Just be aware that there are situations that you might want to change that. e.g. some devices are limited to "baseline" and no b-frames e.g. early ipods (but I would expect the "ipod" preset to reflect that). It depends on what your final format destination goal is

[jagabo]

not that i personally would use it on big outputs because of its cpu usage being 100%

What's wrong with that?

[glenpinn]

hi again, i so appreciate your input, learning something every time i get a reply.

now, without quoting any of your last post, i will just add the following.

regarding Mediaconverter b-frames, it only uses up to 2, so i usually set it to 2 for my outputs when i do use it.

i just did another encode of a 4min 20sec m2ts filmed inside my house not outside and these were the results getting a very close bitrate for each, but the frames in MC are still 30.

Handbrake took 4min 15sec to run
50 frames / RF20 @ 7877bitrate / 244mb / cpu = 60%

Mediaconverter took 75 seconds
30 frames / RF24 @ 7610bitrate / 236mb / cpu = 10% (quick sync is rediculous)

as for the 3770 cpu coping with 100% usage, yeah these things will go all day and night at 100%, and if i had to i would do it, and even with my stock cooler at 100% usage this cpu only hits 65/70 degrees max.

also, its a 3770 not a 3770k so i cant overclock it, i just prefer not messing with these things, im happy with it stock.

What's wrong with that?

no, as i mentioned just above, theres nothing wrong with running these 3770 chips at 100%, what i was referring to was an earlier comment i made where if i have handbrake at 60/70% cpu usage to output, and another program using it at 100% but taking a fair bit longer, i would always use handbrake, especially on longer jobs, but i would use Avidemux on short clips that i had cut from a source file in the time line.

[jagabo]

MediaConverter has screwed up the levels. Darks and brights are crushed. Even with 40 percent fewer fps, and 25 percent more bitrate, MediaConverter is delivering lower quality.

[glenpinn]

MediaConverter has screwed up the levels. Darks and brights are crushed. Even with 40 percent fewer fps, and 25 percent more bitrate, MediaConverter is delivering lower quality.

ok thx, you just helped me deliver the final blow to my licensed copy of mediaconverter, so now i will give my license away to a mate who still uses DVDFab to output his files to h264 AVI, but then again as much as i hate DVDFab for doing any kind of encoding, it may actually be better than mediaconverter.

cheers

[jagabo]

Your 1 minute 40 second source video (in the first post) took about 40 seconds to convert to x264 MKV on my i5 2500K with the veryfast preset, crf=20 (no audio).

[glenpinn]

Your 1 minute 40 second source video (in the first post) took about 40 seconds to convert to x264 MKV on my i5 2500K with the veryfast preset, crf=20 (no audio).

dam you had me worried there for a few moments, thinking your i5 2500k did a 1min 40sec file in 40 seconds ????

then i realized the file was 40 seconds (actually its 37 to be precise)

in handbrake on my stock 3770 cpu it took 40 seconds @ 60% cpu usage (basically real time)

i assume you did it in handbrake, and your i5 2500k is at stock (or is it overclocked) and what usage was it running at ???

i got a bit side tracked the other day and lost the plot regarding your comments in another thread about changing the speed setting in x264 to possibly get handbrake to run quicker.

i still cant get my head around how to check it, and/or change it.

EDIT: did anyone download the handbrake RF20 and RF25 files and check them, i am still deciding if to keep the RF20 setting for handbrake with these m2ts files from my camera, or slide it out to RF23 or RF25, because they would still be pretty dam good.

[jagabo]

Your 1 minute 40 second source video (in the first post) took about 40 seconds to convert to x264 MKV on my i5 2500K with the veryfast preset, crf=20 (no audio).

dam you had me worried there for a few moments, thinking your i5 2500k did a 1min 40sec file in 40 seconds ????

then i realized the file was 40 seconds (actually its 37 to be precise)

Oh, sorry. My mistake!

i assume you did it in handbrake, and your i5 2500k is at stock (or is it overclocked) and what usage was it running at ???

No, I used the CLI encoder, fed by an AviSynth script (ffvideosource()). The CPU was not overclocked.

[glenpinn]

No, I used the CLI encoder, fed by an AviSynth script (ffvideosource()). The CPU was not overclocked.

ok no probs, well you got the same speed as i got with handbrake on a 3770, so i guess in a way, as you mentioned in that other thread yesterday, maybe handbrake already uses the "fast" x264 preset by default.

EDIT: just did the 4min 20sec file again in mediaconverter using CPU only.

took a blistering 1min 40sec to encode (awesome speed for cpu)
cpu = 100%
file size = 226mb
7302 bitrate
frames = 30 (so the frames issue is not caused by quick sync)

EDIT 2: looks like handbrake and avidemux RF20 levels and mediaconverter RF24 all have very similar bitrates.

[glenpinn]

BTW if anyone gets a chance, i still cant work out a small issue im my other thread, the replier cant download the sample file to see what the issue is with handbrake encoding the sample file.

https://forum.videohelp.com/threads/349969-Cutting-AVCHD-m2ts-Files-with-Avidemux

[glenpinn]

Can you post mediainfo (view=>text) on the avidemux encoded file, copy & paste the results back here

will post this up soon

The reason I'm not downloading the other files, is they are not comparable - the bitrates need to be more equivalent. It doesn't really demonstrate anything

ok i see your point now, because before i was assuming that all RF outputs were the same regardless what program i use, and guess i should have known to check the output bitrate for each one first.

i guess the purpose of those files now is not to compare side by side RF20 and RF25 files to each other, but just take a look at the output quality of the Avidemux RF20 & RF25 files and the Handbrake RF20 and RF25 files individually, and offer an opinion about which RF setting is acceptable in each of those 2 programs.

im not concerned about mediaconverter any more so i have taken the links down.

[jagabo]

Since AviDemux and Handbrake both use x264 the RF values mean the same thing. Differences in other settings will will be responsible for the bitrate differences. I usually use crf=18 with a little tuning of other parameters depending on the source.

[glenpinn]

Since AviDemux and Handbrake both use x264 the RF values mean the same thing. Differences in other settings will will be responsible for the bitrate differences. I usually use crf=18 with a little tuning of other parameters depending on the source.

yeah thats what i thought might be the case with those 2 programs, because the RF20 bitrates are pretty close and i try to use output settings as closely matching as possible, and i think i would continue using RF20 in both programs for all my mp4 outputs.

EDIT: i am also looking for a decent external Mic for my camcorder.

https://forum.videohelp.com/threads/350025-External-Mic-for-Panasonic-HDC-SDC750-Camcor...26#post2193026

https://forum.videohelp.com/threads/349994-Panasonic-HDC-SDC750-Cam-%28My-Review%29