FS7: Which software do you use for conversion ?

LennyLevy said:
Doug, are you grading in Resolve before editing? Is the Pro Res output that you use for editing then in 709 with the grade baked in or do you keep it as Log?
If its 709 then you would be limiting your grading options during the cut right? Do you go back to the Log later for final grading? I'm not being judgmental, just curious what people's workflow is as I don't generally edit 4K or Log and I need to up my game eventually with limited power on my machine. Also what graphics card do you have on your Macbook Pro?
Lenny
Click to expand...

My workflow depends on the needs of the production. Sometimes I grade the footage first and bring the exported files (as individual clips with original file names intact) into Premiere. That is my preferred method for short projects or anything that it is scripted and I already know the shots I will use. But other times I edit in Premiere first and then round-trip to Resolve via XML. This is best for a documentary, interviews, etc. when the shooting ratio is high and creative decisions are yet to be made. Both workflows are perfectly valid and have their pros and cons. And even when I grade first, and then later in the edit see something I don't like, I can just re-grade it, export it, and Premiere will automatically replace the old file with the new one. So it is really no big deal to grade first. I prefer it.

The graphics card in my main MacBook Pro is an NVIDIA GeForce GT 750M 2048MB / Intel Iris Pro 1536
 
Last edited:
jpham said:
export to 4096x2160 ProRES HQ, then compress down to H264. How would end user play PRORES - the file is HUGE
what's the time then ? also if you run Mac OS - you have advantage over Windows. Apple never make PRORES codec multithread or support multicore cpu on Windows

Also what the is video bitrate? if you export highly compress - say 30Mbits it sure compress nice and fast
Compress in the 200Mbits VBR and let see how good your machine and codec

Since we're looking for performance number - let make it standard test setup
software - Adobe Premiere
video - XAVC-I - 29.97fps (find a video that has a lot of moving background - ocean waves)
compress to 4096x2160 H264 1-pass VBR to 180Mbps, I-frame every 14 frames
Click to expand...

I don't expect an end-user to casually playback ProRes or ProResHQ. That is my mastering output from which other deliverables can be made later if necessary. BTW, when you export to ProRes the only way to change the bitrate is to choose a different flavor of ProRes, such as ProResHQ or ProRes Lite.

Sorry, I don't have time to go out and shoot 20 minutes of ocean waves just for a speed test. And I don't really care how long it would take to export to H.264 from Resolve because there's no reason I'd ever do it in real life.
 
This thread got me to wondering about the differences in rendering speeds between ProRes and H.264, so I setup a test. The particulars of the codec, resolution, amount of grading applied, etc. really don't matter because the purpose of the test was just to see if, all other things being equal, one rendered out faster than the other. I did two identical exports and the only thing I changed in the Resolve delivery settings was the codec:


5:00 of footage
4096x2160 XAVC-I
S-LOG2
23.98 fps
Grading: 3D LUT applied and a slight reduction in saturation just to make some kind of change to the footage
Exported to "Render at source resolution"


ProRes rendering time: 10:43
H.264 rendering time: 10:57 (quality set to automatic)

ProRes file size: 8.6GB
H.264 file size: 1.4GB
 
Last edited:
Interesting Doug, no transitions, no noise reduction, no sound? Just one clip?

Except for the silly un-European 23,98 fps that is a nice benchmark to test the speed of other systems. And instead of Prores some users will like dnxhq of course, because windows machine. :-)
Does the content of the clip affect the compression? I think so, lots of movement, different colors etc. will have some effect on the compression algorithm I think.

If the content matters, a standard benchmark clip is what we need. I'd like to test it to check my system.
 
Last edited:
Publimix said:
Interesting Doug, no transitions, no noise reduction, no sound? Just one clip?
Click to expand...

Yes, that is correct. Do those numbers sound slower or faster than you'd expect?

I very much doubt that the amount of movement or detail in a shot makes any noticable difference in rendering speeds. I could be wrong, but that is my gut feeling after working with XAVC and Resolve for about four years now. Unless someone can do a test to show otherwise that is what I think is true.
 
Hi Doug, I tested:

5:00 of footage
4096x2160 XAVC-I
S-LOG3
25 fps
Grading: 3D LUT applied

4 clips on timeline, 5.00 min exported to H.264 mpeg4 and H.264 quicktime. PC based computer 12 minutes rendering time. Somehow the render speed varies. file size input unknown (4 clips to get the 5 min.) , output 4,7 GB for both clips. That is a lot bigger than the 1.4 GB from your test. And slower.

Tomorrow I'll do the test again with two 1 min clips, (less time consuming) I will do one clip with moving objects and one static image.

Is there some way I can share these clips so everyone who wants to can get the same clips to benchmark their rendering times?
 
Last edited:
1:00 of footage
4096x2160 XAVC-I
S-LOG3
25 fps
Grading: 3D LUT applied
filesize 1,8 GB

I rendered a 1.00 min clip. with and without slight curve correction in 4K with curve changed rendertime was 135 seconds and without, i t was 114 seconds.

Rendering of a 1 minute clip with moving objects took 94 seconds (4K mpeg H.264 filesize: 412.140)
Rendering of a 1 minute clip with moving objects took 119 seconds (4K DNxHQ 10 bits filesize: 5.702.479)

Moving objects in frame vs no movement seems to make not much difference (as you expected).
 
Last edited:
Always happy to be able to confirm your experience and knowledge. :-)
But just one tiny curve change slows the process with 16%! And the file size from original XAVC-i to DNxHQ doesn't make me want to edit in DNxHD.

Esp. now because I started using Davinci version 14 beta 5, very unstable. I use it like a videogame, constant saving every step.
 
Other software/plugins might be causing the crashes as that's usually the main culprit if your system is fast enough.
 
I'm running 14.0.0b.029 Studio version and like I said, zero problems.
If they have released a newer beta version that is news to me and makes no difference anyway. I'm content just to wait for the final version and I see no need to install every new beta.
 
Doug Jensen said:
I'm running 14.0.0b.029 Studio version and like I said, zero problems.
If they have released a newer beta version that is news to me and makes no difference anyway. I'm content just to wait for the final version and I see no need to install every new beta.
Click to expand...

Publimix said:
I started using Davinci version 14 beta 5, very unstable.
Click to expand...
I am running (testing) 14.0.0b.043 Studio version, and like I said, this version 14 beta 5 is unstable. Even more unstable than version 4 Beta.
But it has some nice new functions.
 
Just curious - some people have written positive things in this thread about Catalyst Browse and Prepare...
Particularly concerning speed of transcoding.

I'm perplexed as this is precisely what I went searching about - my Catalyst Browse is doggedly queuing tasks, processing them one by one, and devoting only one cpu core to the task! It is infuriating - with a 16 core cpu, this just shouldn't be the case in 2018. If multithreaded transcoding of one clip isn't possible for technical reasons related to the codec, the software could at least process several clips in parallel. Also perplexed why the software includes a field to specify the video processing device, and correctly sees my GPU, but gives no option to use the GPU to transcode.

Any help understanding what I am up against would be greatly appreciated!
 
Malcolm Wright said:
Just curious - some people have written positive things in this thread about Catalyst Browse and Prepare...
Particularly concerning speed of transcoding.

I'm perplexed as this is precisely what I went searching about - my Catalyst Browse is doggedly queuing tasks, processing them one by one, and devoting only one cpu core to the task! It is infuriating - with a 16 core cpu, this just shouldn't be the case in 2018. If multithreaded transcoding of one clip isn't possible for technical reasons related to the codec, the software could at least process several clips in parallel. Also perplexed why the software includes a field to specify the video processing device, and correctly sees my GPU, but gives no option to use the GPU to transcode.

Any help understanding what I am up against would be greatly appreciated!
Click to expand...

When I need to transcode raw footage, which is rarely these days, I use 'Edit Ready' from Divergent media which is very fast.
 
morgan_moore said:
Davinci Resolve is free, you can add a LUT - the colours are way nicer than Browse, you can if desired do primary corrections, you can strip off the unwanted audio tracks.
Click to expand...

While I always edit my XAVC-I natively in Resolve, cannot agree that its built-in LUT gives nicer colors; Browse's own LUT is absolutely the best so I even copied it over to my Resolve's LUT directory for those occasions when I use a LUT at all (which is rare as I use RCM or ACES; I also like the color conversion OFX results).

Piotr
 
One solution to slow rendering times from Premiere is nvenc. It's about 6-8 times faster than Adobe Media Encoder for the same h.264 file.
In addition, if you have a Maxwell or newer GPU architecture, it will render out HEVC at faster than realtime speeds. A minute of graded 4K renders h.264 4K in about 30-40 seconds vs 3 mins for AME. HEVC renders in about 45-50 seconds for a 1 minute clip with grading.

What make FS7 footage seem like a breeze to edit is comparing it with high profile h.264 footage from a Phantom 4 Pro 2.0 drone. I have to stop and restart play several times before it will play with some degree of smoothness. Even then, it will drop a few frames every minute. I won't even try recording HEVC on the drone because I don't think CS6 can even import that.
 
You must log in or register to reply here.
Back
Top