the pixelation in foreground is present in the high rez video. encoding will just accentuate the problem.
I’m speaking about the pixelations on the text in foreground. It looks bad.
the pixelation in foreground is present in the high rez video. encoding will just accentuate the problem. I’m speaking about the pixelations on the text in foreground. It looks bad.
Agree about pixelization issue.Nevertheless I just encoded to photo jpeg (quality 99). 2% of white noise added in AE before encoding. http://rghost.net/42180301
Does it looks better for the 13mb/sec size?
Best Regards, Andrey
Okay… I couldn’t resist.
The graphic in the foreground is pixellated because it doesn’t have enough resolution. I guess that you’ve precomped a layer and not collapsed transformations (or if you want transformations uncollapsed, then you haven’t made the precomp big enough).
In terms of the rest of the compression, you’re giving h.264 something that it hates, so yes – in a way your expectations are too high. You have a perfectly uniform color going over to black in a perfectly uniform digital gradient.
What you should do is break the digital gradient up with a fractal noise matte. Keep it at the threshold level just before it becomes perceptible. This gives the digital vignette some shape and stops it being a perfect digital vignette. It would also help to create some texture or color variation in the background… once again, just on the threshold level, so it’s not obviously perceptible, but stops it being a completely uniform colour… anything that breaks up the uniformity will break up the banding when it comes to compression.
Finally, add an adjustment layer with a Noise HLS Auto effect on… settings: Style: Grain, Hue: 3, Lightness: 1.5, Saturation 3, Size: 0.3 and the last one, whatever… it’s just a random seed and render to ProRes / Lossless / whatever.
Then compress this using 2-pass VBR. Start with a bit rate of 14mbits/s or so and move down. Make several experiments with a short clip until you find the optimal data rate / image quality setting for your needs. You will get better results from Quicktime Pro than Adobe Media Encoder. When I have ever run side by side tests, QTPro produces the better results, although AME is quicker.
By the way… it’s h.264 – it will always be a bit blocky. It will never look like lossless. Look a little bit closer at all the previews on Videohive, most have some compression artefacts. You’re probably just being a bit fussy because it’s your own work.
okay very good advice, felt. I will go and fiddle a bit….
okay so I was fiddling all day and rendered it maybe 30 times at least. I got NEARLY there. but it was still BLAH. so what I did was, I shortened the video to 30sec (from 60) to save on the final file size thus enabling me to take advantage of JPEG compression instead of the dreaded H264 and still having a feasible render size. also I fixed the number outlines. rendered to a loseless (are my eyes shot or is it this player, because I’m already seeing artifacts in the loseless) and then to a MOV (JPEG) using qtime pro.http://www.mediafire.com/?dg497hvfmi4y0cv
edit I used the noise idea – very good – cheers
You may very well be able to see banding in the lossless version. This is not because of compression, but because of lack of color depth. Most video formats are 8 bit, and in the case of subtle and smooth gradients, this isn’t always enough colours. Hence you see banding.
One annoying factor is that you may also see banding in 16 bit images… not because there is banding in them, but because you’re looking at them on a screen. And the vast majority of screens display…. yep, you guessed it, 8 bit color. (Some do 10 bit too, but really expensive ones).
So considering that most video is 8bit, most digital files will be compressed at least once some stage in their existence, often quite heavily prior to broadcast, and because most screens that they will ever be looked at are also 8 bit, it’s not actually a bad idea to add little bit of noise/grain to everything you output. Even in an ostensibly uncompressed workflow, it can mean the difference between banding and no banding.
I’ll make a little tut about this at some point, but I don’t have the time right now.
>2 months ago Looks like a long story of noise vs. banding war.
Best Regards, Andrey
Felt sorry for making you busy, but I have something to say about this
I have a Canon 5d mkII, it produces h264 .mov files 38mbps, even if it’s a compressed format I see no color banding at all when I shoot gradients like sky or clouds, I’m desperately trying to match the same encoding settings but I’ve had no luck until now.
The point is: it is possible to make an h264 clip withouth banding, but why we cannot match the same settings of the camera with QT Pro or other software? Do you know any other software? I don’t think so, otherwise you wouldn’t advise to add grain and Noise HLS Auto effect etc.
I understand that there are different solutions, depending on the final products specs, but don’t you think it would be worth to have a look at this? It could solve al lot of problems for me and I believe many other users. I’ll keep searching for this.