I started to produce my UW-videos from a Gopro Hero 12 with Shotcut instead of Adobe Premiere because Adobe don’t support H265. - And for 4K videos there is a large difference of amount of bitrate between H264 and H265 with same quality. -
But I have one problem with Shotcut export: Shotcut do not respond on changing quality value (with hevc-nvenc, quality based(VBR) ) - I get bitrates of about 30-32 mbit/sec in every case -
H265 Sources from Gopro Hero 12 have bitrates of about 50 mbit/sec.
Is this a bug or do I not understanding the proceedings ?
You might be making small changes and expecting a direct correlation, or maybe your video is challenging. We pass a numerical value (cq) to the NVIDIA SDK, and it’s driver and hardware decides what to do.
interesting. I use a NVIDIA GeForce RTX3070 with driver 32.0.15.6094.
Input are mp4 files from the Hero12 in 4K with about 50mbit/sec. Bitrate standard, 8 bit, 60 p/sec
I use the last 2024 shotcut version 24.11.17 - I will do another test
I did an other test: two combined video-files, 52 seconds with cq10, 80% and cq26, 50% - absolutly same result(32mbit/s), no reaction when I change cq !
Do I need Nvidia SDK manager ?
How do you know? Show your evidence. I showed you file size differences in Windows Explorer. I do not know what you are looking at. My source was 4K60 GoPro HERO12 on RTX 4070 on Windows 11. I used the Shotcut export preset “HEVC Main Profile” and then changed the Quality. If you are not using the preset that might be your problem.
I do not recognize the tool you used for your screenshots. Nevertheless, the file sizes are the same. Weird; I don’t know. But if you are concerned with reducing size, perhaps you are more interested in ABR or Constrained VBR modes.
My Screenshots are from the detail view of windows explorer.
I will do another test with my new laptop with RTX 4050 graphics card. I will see, if there I get the same result.
I am not sure, how these video codecs were installed on the pc. So, are there different H265 codecs in use ?
I did an other test on my laptop with the same result. Changing the cq value do not affect the bitrate. The info from avidemux (see after my comment) shows every time the same.
I wil try to understand the “Data of Video Codec” below.
I want to know, why this parameter don’t work for me. - Or do I need this SW-hevc codec from Microsoft Store ? - I think not !?
Shotcut does not use any system-provided codecs except when it comes to the hardware encoder selection, in which case it comes from the GPU vendor.
I am not able to help you further with this issue. If it bothers you enough, in addition to what I already mentioned, you can also not use the hardware encoder.
Hi Dan. You are right, if I do not use the HW acceleration, for example with libx265, i use much longer with full cpu usage for create the video and I get about 2 times bitrate (66mbit instead of 32mbit)
On my 4K screen, I do not see a quality-change when I play it one after the other. I will try to make screenshots from the same moment from original and the two with shotcut created and compare it.
By the way, I created FHD videos from the same source. And also here, the hevc_nvenc codec cq value do not change the bitrate.
I get with FHD about 16 mbit/sec, about half of the 4K.
h265 is really impressive, especially with 4K
It is possible that FFmpeg has a default maximum for some NVENC parameter that we need to override. For example, contrained VBR has both quality and bitrate parameters. Maybe in quality-based VBR where max bitrate is not specified, there is a default in their code that makes it constrained. I will look for that.
I found in a description for the nvenc codec from NVIDIA, that cq=0 means automatic.
When I tried this parm, I get a very small bitrate of less than 5mbit for this 4K video with really bad quality.
That means to me, that there must be an other parameter for setting the best quality, not the smallest bitrate …
I did not find anything like a bad default max bitrate.
Related
52 seconds with cq10, 80% and cq26, 50%
The above link threshold is 24, but your footage might be different if it is more challenging. In other words, you did not try a low enough quality. You are comparing high with higher, and NVENC is not giving you higher because it has some internal limiter. Maybe it worked for you with H.264 because it is less efficient.
I made a crosscheck with Handbrake. Here I took one from my two hero12 testvideos and select the nvenc 4k hw codec and put the quality cq to 20.
The result was the same which I see in Shotcut, about 32mbit/sec.
So, the limiting factor seams to be in the NVIDIA HW codec.
I compared two moments in my test-4k-video between original hero12(about 50mbit/sec) and Shotcut created HW h265 nvenc codec video with about 32mbit/sec - there is no loss of details
