H.264 will always have some banding as it’s a very highly compressed format. Unlike less compressed formats, you can often even see this in naturally shot footage.
Make sure you render everything at 16bit prior to compressing, try to avoid things that will create banding… soft digital gradients, fades to black etc.
Inlife’s is a good suggestion… as is anything else that breaks large gradient areas up. As well as adding noise, you could try adding some texture…. bigger, more structured fractal-noise like patterns at a very subtle level. Nature rarely contains these very smooth gradients and that’s the problem.
Otherwise, it’s a question of dithering with noise / grain. Whereas I can usually find a perfect balance with PJPEG or even MPEG2 , where the noise/grain sits below the perception threshold, but gets rid of the banding completely, this is sometimes not possible with h.264. You will have to choose betweenr some “fizz” from the noise / grain or some banding.
Bear in mind that you’ll probably be way more critical of your own work in this respect. Have a look at my preview for Light Sweep Logo. I had to use noticeable “fizz” to lose the banding. I won’t link it, but you know where to find it. This one was a bit frustrating to me, but at that data-rate, it’s about the best I could do.
Bear in mind also, that not so many years ago all images had “fizz”, ‘cos they were shot on film.
felt_tips saidEmail sent. What is dual compression? I can’t see an option for that.
Mocarg saidSure, go for it. By the way, are you using dual-pass compression?
Felt, can i send you the rendered footage of 20mb to see if the banding level is acceptable?
Inter-frame codecs achieve much of their compression by remembering only the changes between frames. A newsreader in a TV studio will have much more scope for compression than some wobbly hand-held news report shot in a war-zone. If you are compressing the whole news program, a different data rate is suitable for the studio parts than the hand-held footage.
A dual-pass compression analyses the footage in the first pass and assigns bandwidth (within specified limits) according to the image. The second pass does the compression according to that analysis.
Note, that if you do use grain to fool h.264 into not showing banding, you may push the data-rate through the roof. Imagine the news-reader footage in the studio. Only the part of the image with his moving head in it needs to be remembered by the h.264 codec. But if you apply a fat grain to the image, the picture will effectively be changing on every frame and much more bandwidth will be required to compress the footage.
Looks fine to me. There are some compression artefacts for sure, but nothing beyond what you’d expect for h.264 at that compression level. I think you’ve got the balance between grain / banding at a pretty optimal level.
At the end of the day, you’re talking about delivering 7 minutes worth of full HD footage in a 1GB file. That’s no mean feat.
Good conversation and get more knowledge from professionals