You are the french ambassador.
This AMD CPU includes it in the FMA instruction set. FMA can be considered as an extension to AVX. Intel included this set as part of AVX2.
I am curious to know if the windows 7 test version i have sent to you would work for this user?
That might be true, but i want to show an error message at least or write something to the logfile. This can also happen with some NVENC modes. I will think about it and make it nice and clean.
This is not in the FFMpeg part. I do not get 10, 12 or even 16 bit data from premiere. So i need to ask for 32 bit floating point data and then convert it pixel by pixel to the desired bit depth values. Modern CPUs have special vector processing functions to speed this up. In my case this is SSSE3 and AVX2. It seems i need to define some fallback routines for older CPUs.
Found the issue:
I am using the
intrinsic which requires AVX2 and is supporterd by Intel Haswell first (4th Generation).
So older CPUs currently can't make use of 10 bit and 12 bit formats
Yes, it seems the pixel format conversion routine for 10 bit and 12 bit modes does not work on your processor. This is an SSSE3 accelerated routine which normally should work ... but obviousely doesn't. I'll look into it.
So, even with a Windows 7 Pro installation in VirtualBox i can render your example project.
Really, really strange.
- What CPU / Processor do you have?
- Can you try to export with x265 (YUV 4:2:2 10bit)? does this work?
Okay, i will now buy a windows 7 license and install it in a virtual machine.
Okay. I will send you a test-version built for Windows 7 later.
First, the type of the source material is very important. Voukoder does have many renderpaths for bit depth (8, 10 and 12 bit) built in and also converts between pixel formats if neccessary. This is also affecting the encoding speed (i.e. UYVY is much faster than any other format), and this is also part of the reason why it is difficult to compare the encoder speeds). In case of your action camera video it even converts the video from 4:2:0 8bit VFR to 4:2:2 10bit CBR.
But ... it is really strange. I had no problem with rendering your little project - no problems at all.
Unfortunately i have neither Windows 7 nor CC2015, so i can only test with Windows 10 and CC2018.
I assume this is a problem with Windows 7. The current build is targetting Windows 10. I will change this to Windows 8.1 as it is backwards compatible with Windows 7. A direct Windows 7 target is not possible with Visual Studio 2017. But hopefully this will solve the issue.
Thank you for your testing and your error reports. I usually do not get much feedback compared to the high download numbers. It helps me alot. Merci.
P.S.: Voukoder uses the "prores_ks" encoder of FFMpeg. It does not depend on any external library (as far as i know). Can you encode to prores_ks with ffmpeg directly?
Thank you. I have tried your settings and it works fine for me.
Is it possible you send me a small sample project so i can try to reproduce exactly your scenario?
Can you please reproduce it and copy or attach the voukoder.log file here? You will find it in your users "temp" folder.
in the windows explorer and you will find it there.
You are right! Voukoder might need this:
Please try installing it and try again. I think this is included with Windows 10.
Unfortunately i can not test it with CC2015 as this is not installable from the creative clud anymore.
I just fixed this bug. It will work in the next release. Probably 0.7.1.
Currently the user Clockiy is reworking some the Nvidia nvenc parameters. It is a little bit difficult because not all parameters are available for every GPU.
I look forward to see your test results to improve the Voukoder.
The version 0.7.0 ist using the latest git master of ffmpeg from the release day.
You can find information about the parameters here:
I was able to reproduce this issue and will look into it.
Edit: it seems to actually encoder right, but the field order flag is wrong (i.e. in media info). Can you confirm?
Thank you for your feedback! I have checked the file and it should be safe. All it does is changing one byte in GPUFoundation.dll. Hope this helps the CC2017 users. But as far as i know it does not work for everyone.
I have visited your forum (used google translator) and noticed people have some issues with the plugin. It would be great if you could translate that to me so i can solve that issues.
There is nothing wrong. This is how the sample format is stored internally and has to do with data alignment.
There quite some videos about the Voukoder already. This is what i have found so far. Do you know of any other videos comparing or reviewing the voukoder? Add a link to this thread.