![]() So my first suggestion to OP, while him using Handbrake (software libx264b or libx265 implementation), is to try a few different values for CRF and see if the problem persists with significantly lower values than default preset, i.e. You can tune the output bitrate (VBR) with the CQ option of NVEnc in the same way as with the CRF option of libx264. And FFMPEG is a interface to collection of encoders like X265.Being honest, I've never bothered with GUI frontends for ffmpeg so I ignored the existence of Handbrake as a whole just CLI and the good user manual make wondersīTW after some experiments, I've found NVEnc (NVidia hardware-accelerated) encoder to be a few times faster on my hardware compared to libx264, given both similar output video quality and 3-5% larger bitrate (and filesize, respectively). ![]() Learned a whole heck of a lot.ZRGARDNE wrote:Do realize Handbrake, and shutter encoder, are just user friendly GUIs for FFMPEG. ![]() The preset veryslow applies some settings which lets the encoder do a much more thorough analysis on the frames so the encoder can often find better ways to compress the same things, and retain same quality using less bits. So with x264, you could use -preset ultrafast and constant quality 22 and get a 20 GB file, or you could use -preset veryslow and constant quality 22 and a 15 GB file and the video quality would be about the same. It starts with lossless image and then analyzes what stuff can be dropped or blurred or encoded at lower bitrate in order to reduce the quality to level 22. it all depends on the source material and how easy the encoder can achieve that constant quality 22 level.ĭepending on all the other settings you chose, the encoder will use less disk space or more disk space to achieve that CQ 22 level. So one video at Constant quality 22 could be 10 GB, another video at Constant Quality 22 could be 14 GB. In a scene where there's lots of motion, like a car chase or an explosion, it will use more bits to keep quality constant. In some low motion scene like for example just one person talking to another person, the encoder will use less bitrate. You can use more, but some threads will be idle waiting for other threads to finish, or the image quality could potentially degrades a bit, depending on what type of encoding you chose.Ĭonstant quality means just that, the encoder tries to keep the quality constant throughout the video. It’s a front end for FFmpeg and by default it’s using the CPU encoders like x264. It’s very useful and better than Handbrake in lots of ways (e.g. So for a 720p video, around 10 threads would be recommended. Shutter Encoder is a popular free encoding tool so I thought I’d give it a look too. If my memory is correct, for x264 the recommendation was - if you want best quality - to stay with 1 thread per 64-80 pixels of vertical height. Some parts of the analysis of frames ahead can be parallelized, other things can only be done in series. Depending on the settings, the encoder will look ahead at multiple frames and do analysis on those frames, then uses that information and information from previous frames to encode the current frame. The optimal number of threads depends on the settings you use and the resolution of the video. Other interesting open source Windows alternatives to HandBrake are Avidemux, File Converter, VidCoder and Shutter Encoder. In a super simplified way, hyperthreading cores only do some work when there's resources not used by the actual core ( for example let's say the core has 3 "engines" that do additions and 2 "engines" that do multiplications) and the main core works on some code that only uses one engine for addition and the two engines for multiplication - in this case the hyperthreading thread can burrow the unused two engines that do additions to work on that thread instead of sitting idle) The cores created by hyperthreading are not maxed out and that's perfectly fine. According to Code Calamity, AMD's AMF encoder. For reference, NVENC scores 96.13 points, and Intel QuickSync scored 96.37 in this test. Your cpu cores (12 of them) are maxed out. In this benchmark, the highest possible score 100 is points.
0 Comments
Leave a Reply. |