DWA compression

Out of interest, has anyone encountered any issues of note stemming from DWA compression?

You can see the compression in a difference view but I have tried pushing images hard in grade and I visually cannot tell a difference. Anyone experienced any keying issues or anything?!!

To me it seems that mathematically lossless compression is just an easy and safe bet for someone writing a spec doc. In real world terms, there would seem to be much more benefit to working with DWA compression than there are issues. I’d be interested to hear anyone who has encountered problems with it.

1 Like

For CG renders or rendering action outputs, pay attention to the data AOVs (Position, Depth, UV, Crypto, etc). Compression and low bitdepth will affect the results you get out of them.

1 Like

I’d be interested as well. With all the talk from some people here it seems like a wise choice to make the switch.

I’m also curious what people’s thoughts/experiences are with using it as a transport format (i.e. to/from grade) and if there’s anything like generational lose after it being compressed a 2nd or 3rd time.

After a bit of testing of taking the same image in/out multiple times in acesCG, it looks like the only clear thing I notice are rounding differences in extreme highlights where it’s very near pure white. Visually nothing though.

Agree with @milanesa

Used it for 7+ years in commercials I think. Grade no problems to note. Working fully linear/aces and grade at end. All good.

After testing 50 generations on some dark grainy footage we found that it was still visually no different. Under some scrutiny there was compression, mostly around the super bright values. This was mostly values different by a couple of decimal points and certainly nothing that would be obvious without a difference matte test.

This is exactly what our clients do though.

1 Like

For sure @ALan. We’re discussing it as an option for commercial workflows where the clients won’t.

We had a big discussion about it yesterday funnily enough. Doing features & high end episodic work, there are quite often stipulations that you need to work using mathematically lossless or uncompressed workflows. If not for this though, we would definitely be going down the DWA path.

It’s a shame because in regards to infrastructure it makes lots of sense, 10GbE networking is enough to get real time playback from shared storage (if it’s up to it). Smaller shared storage need saves a whole lot of money on infrastructure too.

So even though we won’t use the feature ourselves, I would still advocate for it being a cache & rendering option in Flame. I haven’t done any testing on ST/vector motion maps which would be the one area of concern I would have but apart from that…