help for conversion/resize

[whs912km]

Recode your original 1920x1040 video with this settings

This should compensate the 2.5% overscan and let you see the whole videoframe on your TV.
Change the TV display setting from 16:9 to auto.

oh yes mate! it work!
but in auto output the video is fit to all screen. with output 16/9 finally i see the black bars also on my TV!
but only a small black bars, about 0,5 cm on top and same on bottom. i need to increase the value
also res 1280x692 is too much for my TV (and need more time to re-encode). 540p is enough for me

[jagabo]

The reason ProWo's settings are faster than yours is because is specified the veryfast preset (sacrificing quality for encoding speed). Your command line didn't specify a preset so you reencoded at the default medium preset (sacrificing encoding speed for better quality).

Why don't you just turn off the overscan on your TV? Every modern TV allows you to do that (and by modern I mean since the age of CRTs). It's stupid to reencode all your videos just to compensate for overscan. Every resize loses quality. Every reencoding loses quality. So you lose quality by resizing the video before encoding, and then your TV resizes the video again to overscan. Turning off the TV's overscan would let you avoid all of that.

[ProWo]

The reason ProWo's settings are faster than yours is because is specified the veryfast preset (sacrificing quality for encoding speed).

I don't think so.
From FFmepg h.264 encoding guide https://trac.ffmpeg.org/wiki/Encode/H.264
A preset is a collection of options that will provide a certain encoding speed to compression ratio. A slower preset will provide better compression.
So i think that veryfast gives the same quality as g.e. slow, but bigger files.

[ProWo]

ok mate. i try your setting soon.
another question: for test only (another test ...) i've try to change again the profile level to 3.1 with clever and it work fine very fast
but with simply ffmpeg i've run: ffmpeg -i video.mp4 -c:v libx264 -profile:v main -level:v 3.1 -c:a copy output.mp4
the speed with this code above is very low ... 0.9x
much lower than re-encoding (with -c:v libx264 my speed is 2.1x with 540p video resizing)
how your app is very fast? which string does it use for simple profile switching?

Because your commandline reencodes the video and my app doesn't.

[ProWo]

Recode your original 1920x1040 video with this settings

This should compensate the 2.5% overscan and let you see the whole videoframe on your TV.
Change the TV display setting from 16:9 to auto.

oh yes mate! it work!
but in auto output the video is fit to all screen. with output 16/9 finally i see the black bars also on my TV!
but only a small black bars, about 0,5 cm on top and same on bottom. i need to increase the value
also res 1280x692 is too much for my TV (and need more time to re-encode). 540p is enough for me

For 540p output and correct Video AR use:
Frame Width 956
Frame Heigth 516
Pad Width 960
Pad Height 540
DAR must shown 1.78

If you want increase the black bars (but this alters the Video AR, not recommended), then use:
Frame Heigth 502 (Video AR 1.9:1)
or
Frame Heigth 478 (Video AR 2:1)
The other values as above.

TV display setting to 16:9

[whs912km]

jagabo, you right. but on my TV isn't possible to turn off overscan. this is my real problem ...

ProWo
your commandline reencodes the video and my app doesn't
yes, i know that. but if i leave -c:v libx264 from my code above, ffmpeg use by default libx264 ...
so, my question is: how to change level profile on ffmpeg without reencoding like your app??

about your setting for my TV i can try now. thanks

[jagabo]

So i think that veryfast gives the same quality as g.e. slow, but bigger files.

I know that's the mantra. But in my experience some of the slower presets (faster, fast, sometimes medium) usually give higher bitrates than veryslow. Then bitrates usually get progressive lower with slow, slower and veryslow. Placebo often gives slightly higher bitrates than veryslow. In most cases the difference in bitrates between veryfast and the slowest presets is less than 10 percent. Of course, this is with CRF encoding.

But visual quality also generally improves as the presets get slower. This is because some of the setting (me, subme) result in better quality and higher bitrates, while others increase the amount of compression.

[whs912km]

hi guys, here my results:

for me (on my TV), the best setting is
Frame Width 956
Frame Heigth 456
Pad Width 960
Pad Height 540

speed:
with veryfast 3x
with faster 2.1x
with fast 1.6-1.7x (but the quality on my tv os the same. so for me veryfast is better)

now ProWo only a little problem: on my converted mkv video, no audio ....! why??

[jagabo]

jagabo, you right. but on my TV isn't possible to turn off overscan.

Unless your TV is a CRT I doubt that's the case. Keep in mind that it's not called "overscan" in the TV's settings. It goes by names that vary from manufacturer to manufacturer (maybe even by models). Look for "pixel-for-pixel", "just scan", "perfect scan" etc.

[ProWo]

now ProWo only a little problem: on my converted mkv video, no audio ....! why??

Because you have processed the video only.
Reload your original file, click main.click multiplex.
Drag and drop the new created video.
Control now all the streams you wish to mantain. This streams must be checked on the left side, the other streams unckecked.
The new created videostream must be checked of course, the original videostream unchecked.
Then control if the container is right and click multiplex.
Done.

[whs912km]

i don't have nothing on my tv about this setting.

anyway, with ffmpeg my final working code (for my video and my tv) is:
-acodec copy -c:v libx264 -crf 23 -preset:v faster -vf scale=960:456,pad=960:540:0:48

this is all the best for me (quality, speed, resolution)

@ProWo
ok, understand. now multiplex work for me. final video+audio ok. thanks

[jagabo]

i don't have nothing on my tv about this setting.

If your TV really doesn't have a non-overscan mode -- many media players have an underscan option which does what you want while playing the video, obviating the need to add borders and reencode.

[zing269]

Assuming that this is the correct manual for the tv, https://www.telesystem-world.com/uploads/attachments/10-25-2011-manuale-tv-led-palco19...ed01-10000.pdf, then there doesn't appear to be a setting for overscan. It may be that only the PC dsub15 vga connector has overscan disabled.

[jagabo]

Sometimes setting an HDMI input to "computer" will disable overscan. Or one of the HDMI inputs may be dedicated to computer use -- and overscan will be disabled.

[Sharc]

anyway, with ffmpeg my final working code (for my video and my tv) is:
-acodec copy -c:v libx264 -crf 23 -preset:v faster -vf scale=960:456,pad=960:540:0:48

With this scaling of 960:456 you actually convert your square pixel 1920x1040 source to an anamorphic picture with a Pixel Aspect Ratio of 0.877 which is pretty close to a standard 4:3 NTSC DVD (which has a Pixel Aspect Ratio of 0.889). Then you pad it to 960:540 pixels (=1.78=16:9). It looks like your playback scenario plays it back accordingly, means according to the anamorphic Pixel Aspect Ratio of 0.877 which returns a full frame (i.e. including the borders) of 842:540 pixels with a correct active movie picture of 842:456 pixels = 1.85:1.

Edit:
Out of curiosity using your commandline, this is how my 16:9 TV plays it boxed, with top+bottom bars from your padding, and left and right pillars padded by the TV.
The movie is displayed correctly at 1.85:1 aspect ratio. Your TV's 2.5% overscan will definitley not hide anything of the movie.

[whs912km]

thanks for your clarification, but i don't understand when you says Your TV's 2.5% overscan will definitley not hide anything of the movie

whit my code above i can watch fine my video in 16/9 with black bars. and that's what i wanted since my first post of this thread ...

from my side question is solved

[Sharc]

... from my side question is solved

Which is what really counts