Better To Put Brightness, Contrast, & Volume Up A Bit?

[video_magic]

In my Newbie considerations I have settled on the above opinion after watching a lot of MPEG1 & MPEG2 videos.

It's all about getting the detail.

I have often found it preferrable to watch a slightly brighter MPEG1 clip, than a higher-bitrate, but darker MPEG2 encode!

This is because darkness seems to be the main spoiler and hider-of-detail in MPEG from what I can tell. Things in the background are 'lost' in darkness, but on clips that were encoded a little brighter I can then see them! There is little more annoying than clips that are too dark - and not being able to make out the details in the video.
Throwing high-bitrate at dark sources doesn't seem to be the answer - I suggest that turning up the brightness by a small amount on your input before encoding to MPEG is the way to go! :)
Obviously not too bright, as this will negate the good work and lose the details in 'liight' saturation.

On the same angle of making out the detail, I have found that it is better to have louder-encoded audio at a lower-bitrate, rather than quiet audio. It seems that encoding from a louder source (i.e. with the audio turned up a bit on the input) results in being able to hear the detail more and it being enjoyable to listen too. Much better than encoding quietly and throwing high-bitrate at it.

So, I think I am correct with these findings:
(1)Up the brightness 'a little above normal' when capturing video, before encoding it, rather than have it dark.

(2)Up the volume a small amount likewise, rather than have it too quiet.

I think the results will be more pleasing to the viewer even if it were at a slightly lower bitrate :)

[Mug Funky]

hmmm. i've wondered about whether feeding the encoder with gamma-corrected frames for ratecontrol decisions, but actually encoding straight frames would be a good thing. it's not trivial to test out though...

one thing though - if details are getting buried, try turning your TV's contrast to full, and it's brightness to just above half-way. this might seem harsh to begin with, but it's a good way to monitor because nothing is being lost in the TV.

about audio - volume _should_ be non-negotiable, at least in theory. the broadcast standard levels vary a bit, but generally it's -20dB down from fullscale (digital) is equal to 0 dB VU (analog level). and peaks should be at -10dB FS (10dB higher than the reference level).

this is rarely true though. i've found japanese audio is often louder than it's english counterpart (in anime of course), but older japanese stuff that's been dubber recently the opposite is true (dragonball, YYH are good examples).

as far as encoder performance, something as simple as global gain shouldn't throw off a psy-model too much, but sadly it does quite often (even in the best encoders - like LAME's performance on extremely quiet classical samples, as demonstrated by guruboolez on HA.org). mp2 encoders probably suffer this a fair bit more, but ac3 is pretty good with it - it was after all designed from scratch to work on broadcast sources like movies etc that are at a common reference level. then again, mp2 has this in common with it...

[video_magic]

Thanks for your input Mug Funky.

There are so many things I'm just finding out. I have been working with material from lots of different sources, and I would like a common reference - I mean I don't want to keep altering an output display to make decisions like whether a pic is too bright. But I will take your advice on how to setup the output picture - and then leave it like that - then try matching up all my encoding pictures based on it. I guess that is as close to setting everything to around a reference as I could hope for! :)

At this time I am still of the opinion that a little brighter than 'default' is better than to be dark - I know this is altering the picture from some reference standard, but it seems better for the 'enjoyment' of the viewer.

With audio are you saying it is best not to louden it before encode? I should be encoding to ac3.

Thanks a lot for your advice, I am struggling to make the right decisions as I get further in to all this :)

[Mug Funky]

well, ac3 has dialog normalization and a setting for mix level that affects the way it's played back (if DRC is turned on on the playback device that is).

if the values aren't filled out correctly (which is pretty much always the case... over here we just leave it at -27dB dialnorm unless the tape actually says something), you can get pumping in playback, or too quiet/loud (usually too loud) audio, which can be very annoying.

just for the record, people who master for music DVDs do not have a clue what they're doing - they'll master it like a CD, which of course means it'll be skull-crushingly loud when played back on a system set up for movies (which are mastered about 15dB lower... that's a huge difference). usually i turn their stuff down without telling them... foobar2000's replaygain will set things at the right level, so what i often do is scan a wav with replaygain, then output it with diskwriter to 32-bit float with replaygain turned on. then i can safely encode at -27dB with 89dB mix level and know those are the right values. thankfully Soft Encode supports 32-bit float wav files... 16 bits is okay, but i don't like to lose anything from a simple gain adjustment.

as for video levels, there's not much of a standard there. just so long as it's between 16 and 235 luma, there's no problem - anything goes.

i agree that a too-dark movie can be quite annoying. often the director wanted it that way though, so i tend not to adjust video levels unless it looks like crap and needs it (usually if there's a problem like that with video, it's the encoders and authors that get blamed, rather than the guys who made it, so it makes sense we should fix it).

[Milo Bloom]

Several years ago when I used to exclusively use TMPGENC I found that for most videos the uping the brightness often created an encode that seemed to look superior to the original. However, in very dark scenes and in scene transitions (especially those that go to black) the results were unsatisfying as artifacts would appear for a frame or two (you really couldn't tell unless you were watching for it) as well as the screen would be slightly gray where it should be black. I now leave the brightness alone.

[Mug Funky]

hehe. conversely, if you're transcoding from an already lossy source, it may help matters to actually crunch out the darks - better to have nothing at all than a sea of flickering blocks... but i wouldn't recommend doing this.

[edit]

i tend to think of video levels as being analogous to audio levels, except at best video only has 10 bits of dynamic range (~60dB) and usually only 8 bits (48dB). audio has a much higher dynamic range, and is more amenable to dithering (which will exploit human hearing to jack the dynamic range up from 96dB to almost 110!), so we can afford to master at lower levels for audio.

it's good practice for video to use up as much of the range as possible without clipping (washed out whites or crushed blacks), and without disturbing the intent of the video - the director might have wanted it dark, and an encoder should respect that.

no dynamics means boring video and boring audio... i guess it puts the focus on the acting, writing and directing. with the state of the movie industry, the acting writing and directing needs all the help it can get :)