Just wanted to say thanks -- I've been using Handbrake for over a decade, but defaults and the occasional experimentation was fine. Was unaware of hardware encoding. And was getting really fed up with all the youtube videos that only make big Handbrake claims, but basically just show you how to download it, pick a preset, and hit go. Your video is very in-depth, insightful, and useful.
Amazing video, you clearly understand the topic and are not just throwing information you don’t know about like 90% of the RUclipsrs that covered this matter.
Omg thanks a lot for this comparison. I have been looking for such a scientific approach for hours and was even up to setting up vmaf on my own to check the best setting and especially the comparison between NVENC and CPU. Thanks a lot for saving me lots of time! :)
Great video. It's made me rethink my compression work flow. Previously I used x265(CPU) at med/24 settings, but I was shocked that NVENC (Turing, 6th Gen) could produce almost the same file size, at 1/4 the speed and score 1 VMAF point higher than the CPU.
Yeah there is a LOT of anti-GPU sentiment among the home streaming/server community, if you ever ask about NVENC around them you'll get crucified for even daring to mention anything except H265 CPU on Slower preset, but I agree with the video. The quality difference isn't THAT big of a deal in my eyes, doing A/B testing between myself and my brother to see if we could tell CPU (QC 20, Medium) versus GPU (QC 20, Fast) and we were getting basically 50/50 with our guesses. We also saw around 94-98 VMAF scores depending on the quality of the source (TV series, especially older ones, tend to already be kinda blurry on bluray so you get way more leniency on compression). I just can't justify wasting time CPU encoding anymore. My GPU is going like 5x faster and that really matters when you want to put a 10 season long TV series onto your home server.
This is some great investigative work and editing! Super useful. I'm not sure if NVENC has evolved to the point of delivering the same compression ratio as h264/h265 however. When I tried it last a couple years ago, the quality was relatively poor and the file size was not even worth compressing at all. I'll try compressing a video with your settings though, now that I have an RTX card. Maybe things will be different
I have a follow up video. even though nvidia 1000 series (1060 1070 1080 etc) does have nvenc, it is incredibly basic and will not offer compression and quality results shown in this video. I used the 2060 in this video and assumed that all nvenc performed equal which was a lapse on my part. different generations of cards have different nvenc "qualities" and with a newer card, will give better video compression results with smaller file sizes.
It works weird, same configuration on a 7gb 4K movie give me a result of half gbs with great quality, around 3.7gb. Yet I tried a 6.7gb 4K movie and the result was 6.3gb... such a weird, welp, my card is a 1050ti so maybe I shouldn't have expected more XD.
@@axell.2875 because its depends on your card generation PASCAL DOES NOT SUPPORT HEVC H265 b frames, only from turing support, so you can guess files size and quality without b frames lol. So on pascal its better using nvenc h264 only i think since there bframes support to 4 frames.
The video was absolutely breathtaking! The information is very useful, and your time & effort is greatly appreciated! I'm going to leave a link in my blu-ray video to this video for compression advice!
INCREDIBLE video my friend you literally showed what each of them do and also showing comparisons and graphs of how much difference they make unlike others who just say put it to 23 quality and let her rip😆🤣
Well I probably won't re-encode my entire library with h265 understanding that moving forward will make a huge difference in what I choose to use. Thanks for the great information!
I realize it would probably be a huge time burner (depending on how big your library is) but it would probably allow you to shrink the storage space by half and retain the same quality... which may or may not be worth your time. At least there's batch processing with Handbrake, and I noticed that according to Task Manager, one encode wasn't utilizing the GPU's video encoder but about 40% (though the CPU was pegged) so I decided to set it to do 2 simultaneous encodes. With the newest NVIDIA drivers, even more may be possible. I've got a lowly GTX 1050 and an i5-4590. Literally the lowest card able to do NVENC and it can handle two simultaneous encodes without trouble. I can only imagine what a newer system could do.
@@SeeJayPlayGames actually handbrake wouldn't be the bottleneck for me. The GPU encoding goes pretty quick. It's the fact that I would have to extract all the physical media into digital media that would be the bottleneck and then encode it to the right settings
@tor13128 You can look the graph 4:49, the H265 Ultrafast got around 98 VMAF, whereas H264 Fast lower (96.7), and even worse, H264 at Ultrafast the VMAF as low as 88. That’s why the recommended setting is H265 - Ultrafast. Not much of decrease in quality.
@@danielwebb9621 thank you for replying. He is talking about using CPU -> H265 -> Ultra Fast giving better quality than all H264 combinations? 4:53 Which of these screenshots are H264 and which are H265? They look very bad at Ultra Fast.
This video is extremely helpful! Thank you so much! Question: If Using Ultra Fast speed gives the worst quality why are you saying that we should use it?
What also can be interesting for people trying to save the clips in as good of a quality as possible, without using up your Disk space is this: I'd like to my videos upscale to 8k (nearest neighbor) using ffmpeg and then upload it to youtube. Why 8k? Because it allows youtube to retain more quality because of the bitrate, and it also enforces youtube to use the AV1-Codec, which means more quality for the same bitrate or same quality with less bitrate. It works well with rec709 stuff, but It doesn't work on Videos with Log-Profiles. You'd have to colour grade your footage to make it work. I also use my Laptop GPU (RTX 4070 Mobile) to encode it directly to NVENC_AV1 (still at around 200mbit bitrate though), in order to get very good quality, smaller filesizes than with other codecs to reduce upload time. I also use a custom script to iterate through an entire folder including subfolders to re-encode videos. Now i just have to figure out how to upload them without actually being at my computer:
I would love for you to test using Xmedia Recode and Two-Pass VBR encoding both CPU and GPU. Xmedia Recode is the only freeware I know of that lets you do Two-Pass NVENC VBR encoding.
hmm, I've been playing around with handbrake settings, and my movies are coming out at 10-12gb using HD passthru for audio with the quality slider at 18. For my tests I used a blu ray 1080p rip of puss n boots and the emoji movie, which are 20ish GB in size. I didn't see an appreciable difference in file output size between setting the quality slider at 18 vs 24. .264 did seem to encode 2 or 3 minutes faster than .265, I used NVENC for both with a spare gtx 1050 2gb gpu.. all using the slow encode speed. average encoding time for those movies was anywhere from 15-19 minutes.
Good but using Gpu gets bigger files with h265 i tested and i saw differences using 264 vs 265 and 264 get smaller files not much but it's something compared to 265.
Superb video, so I have it in my head to want to put the quality slider to 0 (lossless) and thinking I get the best quality available, with slow encode speed. Is that not best practice?
Any recommendations for game footage compression? It's natively recorded in x265, but the file size is ridiculous. I lowered the bitrate of the recordings to around 7000-12000 (I could go lower, but am afraid quality will suffer too much at that point), but still chose to later encode with the quality slider at 35(!) to get more sane file sizes relative to recordings times, even realizing that recording in 60 FPS and keeping the quality bumps the size up (thought probably not to exactly twice the size of a 30FPS video) it still was just too much.
Depends. If you plan to store the files after editing/uploading then use softare h265 on slow with cr at 20/22/24. It will be slower, but the quality and size will be improved significantly. However, if the footage is going to be trashed after a video you might as well just edit the video and render using nvenc to save time as its going to be transcoded when you upload it anyways. Use CR/CF over bitrate.
@@VelocityFTW I ended up just doing the safest thing: encoded some clips as samples in many, many, MANY, different settings and ended up with the conclusion that the recommendation was on point. Veryfast on 22~ (depending on how intricate the source is, since some levels in the game are incredibly color-pallette heavy while some are muted) seems to yield the best compression to accuracy ratio
@@VelocityFTW Software. Took about two weeks to finish all experiments. I did quite a lot, and on footage from multiple games both 2D and 3D to get a general idea of overall good settings for videogame footage compression with fair quality. But I can also tell you that some games benefit immensely (size-wise) from specific settings while losing negligible quality, so my recommendation is to take 3-4 clips 3-10 minutes long and test what works best for that game's footage. Of course, if you don't stand to gain significant savings from this it'll be a waste of time, but I went down from 130GB to 80GB with barely any perceptible quality loss, so for me it was definitely worth it.
questions. I'm trying the h265 nvenc with handbrake, using same settings and same videos(tested with multiple videos). the only difference is the resolution. if i set the output 1080, it will utilize my gpu as per intended with nvenc. but if i set the output to 480p it will goes back to cpu encoding. any idea why and how do i get 480p to use my gpu?
im trying to compress 30 second clips of 2560x1600 in 60fps and dont care about the quality i just have to kinda be able to tell whats going on in the video while its still 60fps ive managed to compress it from 35mb per video to 7mb but do you think i could shrink the file size even more?
if anyone in the future watches this AV1 will allow you to drop to 36 from 30 with darn near lossless footage - i had a 15.5GB file in h.265 compressed with AV1 3.36GB couldn't tell the difference with quailty AV1 is 1000times better than h265
hardware encoding is good at handling digitally recorded video sources. but when it comes to the analog sources with film grains... software encoding wins
Never ever use MP3 for anything. It's been outdated since the late 90s. AAC has the same compatibility and much higher quality. For extreme size savings OPUS is a great choice too
Yea I initially wanted to also cover different types of videos like online lectures, teams recording etc but there were too many permutations and edge cases to cover.
Hey, great Video! really in depth an informative! bravo. I have one question. Are your quality recommendations resoulution dependent? I want to convert 1080p and 2160p material with nvenc. Thank you in advance! @Anitomicals
Using the quality slider is resolution independent. Files will come out with approximately the same quality with the same slider value regardless of resolution. (but of course it will end up at higher file sizes at higher resolutions) To be resolution dependent, there is an option below the slider called AVG BITRATE (KBPS). input a number, usually from 80,000 for 4k and 20,000 for 1080p. it will hard cap your final file size to a desired size and hence will affect the output quality with different resolutions if you change the number. I picked the slider number exactly for this reason, I dont want people to fuss with different sizes at different resolutions.
@@Anitomicals thank you a lot! This clears everything up for me. This logic makes sense, and I am happy that they implement it like that. I had this thinking because of the info text in handbrake when hovering over the quality slider. But it says it’s for h264 not h265. thats why I was asking in the first place to be sure about h265 :)
Just wanted to say thanks -- I've been using Handbrake for over a decade, but defaults and the occasional experimentation was fine. Was unaware of hardware encoding. And was getting really fed up with all the youtube videos that only make big Handbrake claims, but basically just show you how to download it, pick a preset, and hit go. Your video is very in-depth, insightful, and useful.
Amazing video, you clearly understand the topic and are not just throwing information you don’t know about like 90% of the RUclipsrs that covered this matter.
Brilliant video; excellent analysis with clear and straight forward explanation. You've saved me a lot of time and SSD space! Cheers!
My goodness... Thank you for this gem. Commenting to help the RUclips algorithm recommend this fantastic tutorial!
Nice! Finally a good tutorial about Handbrake and compression. Good work! And thanks!
bro qualities of your videos are great in all aspects.
I was struggling to get good file size using NVENC, you solved it. Thank you
Hi; How have you solved it? I still dont get such smale files.
@@RoFBOPE same have you solved it 4 months later?
this is the most informative video i have ever seen on video compression, amazing work, love it.
Omg thanks a lot for this comparison. I have been looking for such a scientific approach for hours and was even up to setting up vmaf on my own to check the best setting and especially the comparison between NVENC and CPU. Thanks a lot for saving me lots of time! :)
Great video. It's made me rethink my compression work flow. Previously I used x265(CPU) at med/24 settings, but I was shocked that NVENC (Turing, 6th Gen) could produce almost the same file size, at 1/4 the speed and score 1 VMAF point higher than the CPU.
Yeah there is a LOT of anti-GPU sentiment among the home streaming/server community, if you ever ask about NVENC around them you'll get crucified for even daring to mention anything except H265 CPU on Slower preset, but I agree with the video. The quality difference isn't THAT big of a deal in my eyes, doing A/B testing between myself and my brother to see if we could tell CPU (QC 20, Medium) versus GPU (QC 20, Fast) and we were getting basically 50/50 with our guesses. We also saw around 94-98 VMAF scores depending on the quality of the source (TV series, especially older ones, tend to already be kinda blurry on bluray so you get way more leniency on compression). I just can't justify wasting time CPU encoding anymore. My GPU is going like 5x faster and that really matters when you want to put a 10 season long TV series onto your home server.
This is some great investigative work and editing! Super useful.
I'm not sure if NVENC has evolved to the point of delivering the same compression ratio as h264/h265 however. When I tried it last a couple years ago, the quality was relatively poor and the file size was not even worth compressing at all. I'll try compressing a video with your settings though, now that I have an RTX card. Maybe things will be different
Just tested this and using NVENC results in files with three times the size of the ones generated via cpu using a GTX 1070. No NVENC for me then.
I have a follow up video. even though nvidia 1000 series (1060 1070 1080 etc) does have nvenc, it is incredibly basic and will not offer compression and quality results shown in this video. I used the 2060 in this video and assumed that all nvenc performed equal which was a lapse on my part. different generations of cards have different nvenc "qualities" and with a newer card, will give better video compression results with smaller file sizes.
@@Anitomicals Ironically, the H.265 ultra fast 30.. comes off as a pixel mess. But the NVENC on my GTX 1070 comes of perfect and pretty small.
It works weird, same configuration on a 7gb 4K movie give me a result of half gbs with great quality, around 3.7gb.
Yet I tried a 6.7gb 4K movie and the result was 6.3gb... such a weird, welp, my card is a 1050ti so maybe I shouldn't have expected more XD.
@@axell.2875 because its depends on your card generation PASCAL DOES NOT SUPPORT HEVC H265 b frames, only from turing support, so you can guess files size and quality without b frames lol. So on pascal its better using nvenc h264 only i think since there bframes support to 4 frames.
Great video, Im using always nvenc h265 on slowest, because its pretty fast enough
To the point and very concise. Thank you! Love this video
Thank you very much for sharing all your testing on this!
The video was absolutely breathtaking! The information is very useful, and your time & effort is greatly appreciated! I'm going to leave a link in my blu-ray video to this video for compression advice!
This was surprisingly thorough. Well done
Very nice dude =). I use handbreake a long time... this is star info.
INCREDIBLE video my friend you literally showed what each of them do and also showing comparisons and graphs of how much difference they make unlike others who just say put it to 23 quality and let her rip😆🤣
This is the best video explain on youtube thanks man.
Wow thank you so much for this incredible detailed video! Explained really well and down to the point, perfect!
An excellent and very concise description of the h265 encoder options.
This video is pure gold. Thank you very much!!!
Great video! Clearly explained! This must have taken a lot of research and trail&error.
WTF THIS IS EPIC NICE DUDE
Fantastic - Informed - No BS
This is exactly what i was looking for. Thank you.
thx for this in depth vid. have a good day.
Thani you so much. Now thar my RTX 4080 is doing it, its taking 1 hour instead of 6. Crazy! thank you so much!
Great video, thank you!
Very helpful, thank you
I can use NVEnc but with cpu it compresses much better when i tested. It was about 50% smaller with the same quality preset
Very intructive video. Thanks for your explanations!
Well I probably won't re-encode my entire library with h265 understanding that moving forward will make a huge difference in what I choose to use. Thanks for the great information!
I realize it would probably be a huge time burner (depending on how big your library is) but it would probably allow you to shrink the storage space by half and retain the same quality... which may or may not be worth your time. At least there's batch processing with Handbrake, and I noticed that according to Task Manager, one encode wasn't utilizing the GPU's video encoder but about 40% (though the CPU was pegged) so I decided to set it to do 2 simultaneous encodes. With the newest NVIDIA drivers, even more may be possible. I've got a lowly GTX 1050 and an i5-4590. Literally the lowest card able to do NVENC and it can handle two simultaneous encodes without trouble. I can only imagine what a newer system could do.
@@SeeJayPlayGames actually handbrake wouldn't be the bottleneck for me. The GPU encoding goes pretty quick. It's the fact that I would have to extract all the physical media into digital media that would be the bottleneck and then encode it to the right settings
true, true... takes a while to rip a BD I suppose @@flythereddflagg
4:55 Am I misunderstanding the video? How come it looks lower quality at ultra fast but the recommendation is to use ultra fast?
@tor13128 You can look the graph 4:49, the H265 Ultrafast got around 98 VMAF, whereas H264 Fast lower (96.7), and even worse, H264 at Ultrafast the VMAF as low as 88. That’s why the recommended setting is H265 - Ultrafast. Not much of decrease in quality.
@@danielwebb9621 thank you for replying. He is talking about using CPU -> H265 -> Ultra Fast giving better quality than all H264 combinations?
4:53 Which of these screenshots are H264 and which are H265? They look very bad at Ultra Fast.
Wow. Amazing video with scientific methods.
this is very informative. thank you!!
good job, don't have anything to say except it's well researched and edited actually ty im using a laptop
This video is extremely helpful! Thank you so much!
Question: If Using Ultra Fast speed gives the worst quality why are you saying that we should use it?
Amazing video, but could you make one for av1?
Thank you this is gold info!
I still find H264 best especialy when you have legacy devices and new devices that still haven't adopted H265.
What also can be interesting for people trying to save the clips in as good of a quality as possible, without using up your Disk space is this:
I'd like to my videos upscale to 8k (nearest neighbor) using ffmpeg and then upload it to youtube. Why 8k? Because it allows youtube to retain more quality because of the bitrate, and it also enforces youtube to use the AV1-Codec, which means more quality for the same bitrate or same quality with less bitrate.
It works well with rec709 stuff, but It doesn't work on Videos with Log-Profiles. You'd have to colour grade your footage to make it work.
I also use my Laptop GPU (RTX 4070 Mobile) to encode it directly to NVENC_AV1 (still at around 200mbit bitrate though), in order to get very good quality, smaller filesizes than with other codecs to reduce upload time. I also use a custom script to iterate through an entire folder including subfolders to re-encode videos.
Now i just have to figure out how to upload them without actually being at my computer:
Have you found a way to change blocksize with nvenc? The quality slider wont really matter after 24 unless you can right?
Thanks, awesome video!
More than excelent perfect
I would love for you to test using Xmedia Recode and Two-Pass VBR encoding both CPU and GPU. Xmedia Recode is the only freeware I know of that lets you do Two-Pass NVENC VBR encoding.
Great Video! Can you repeat this with AV1 and Fastflix (FFMPEG)?
you are a legend
ultra super thanks for you
hmm, I've been playing around with handbrake settings, and my movies are coming out at 10-12gb using HD passthru for audio with the quality slider at 18. For my tests I used a blu ray 1080p rip of puss n boots and the emoji movie, which are 20ish GB in size. I didn't see an appreciable difference in file output size between setting the quality slider at 18 vs 24. .264 did seem to encode 2 or 3 minutes faster than .265, I used NVENC for both with a spare gtx 1050 2gb gpu.. all using the slow encode speed. average encoding time for those movies was anywhere from 15-19 minutes.
Thank you that was most helpful and thorough! Now in handbreak there are 2 options for cpu H.265: 10 bit and 12 bit. Which one would you recommend?
great video
thank you very very much.
Truly the best indeed
Good but using Gpu gets bigger files with h265 i tested and i saw differences using 264 vs 265 and 264 get smaller files not much but it's something compared to 265.
Superb video, so I have it in my head to want to put the quality slider to 0 (lossless) and thinking I get the best quality available, with slow encode speed. Is that not best practice?
Only the best
Any recommendations for game footage compression? It's natively recorded in x265, but the file size is ridiculous. I lowered the bitrate of the recordings to around 7000-12000 (I could go lower, but am afraid quality will suffer too much at that point), but still chose to later encode with the quality slider at 35(!) to get more sane file sizes relative to recordings times, even realizing that recording in 60 FPS and keeping the quality bumps the size up (thought probably not to exactly twice the size of a 30FPS video) it still was just too much.
Depends. If you plan to store the files after editing/uploading then use softare h265 on slow with cr at 20/22/24. It will be slower, but the quality and size will be improved significantly. However, if the footage is going to be trashed after a video you might as well just edit the video and render using nvenc to save time as its going to be transcoded when you upload it anyways. Use CR/CF over bitrate.
@@VelocityFTW I ended up just doing the safest thing: encoded some clips as samples in many, many, MANY, different settings and ended up with the conclusion that the recommendation was on point.
Veryfast on 22~ (depending on how intricate the source is, since some levels in the game are incredibly color-pallette heavy while some are muted) seems to yield the best compression to accuracy ratio
@@nullfield1126 Using nvenc or software?
@@VelocityFTW Software. Took about two weeks to finish all experiments. I did quite a lot, and on footage from multiple games both 2D and 3D to get a general idea of overall good settings for videogame footage compression with fair quality.
But I can also tell you that some games benefit immensely (size-wise) from specific settings while losing negligible quality, so my recommendation is to take 3-4 clips 3-10 minutes long and test what works best for that game's footage. Of course, if you don't stand to gain significant savings from this it'll be a waste of time, but I went down from 130GB to 80GB with barely any perceptible quality loss, so for me it was definitely worth it.
Most other channels are claiming h.265 is a non starter due to licensing issues and just talk about how awesone AV1 is compared to h.264
No bullshit straight to the point. Just awesome
So I did it with the method you show us but from 61 mb short move file it became 87mb
i've got a probl the n vidia encode never exist on my handbrake
questions. I'm trying the h265 nvenc with handbrake, using same settings and same videos(tested with multiple videos). the only difference is the resolution. if i set the output 1080, it will utilize my gpu as per intended with nvenc. but if i set the output to 480p it will goes back to cpu encoding. any idea why and how do i get 480p to use my gpu?
could you make a tutorial on how to use vmaf?
im trying to compress 30 second clips of 2560x1600 in 60fps and dont care about the quality i just have to kinda be able to tell whats going on in the video while its still 60fps ive managed to compress it from 35mb per video to 7mb but do you think i could shrink the file size even more?
Great. What's the best setting for AV1 code.?
Liked, shared & subscribed
I don't have the hardware preset, am I missing something?
GPU is 1060 gtx.
Now it works has to reset presets, chers, and nice video
if anyone in the future watches this AV1 will allow you to drop to 36 from 30 with darn near lossless footage - i had a 15.5GB file in h.265 compressed with AV1 3.36GB couldn't tell the difference with quailty AV1 is 1000times better than h265
thats exactly what i am looking for i am about to delete my orignal files and want to be sured that my cq 30 is good or not, that you verry much
Where can I find Ur VMAF-Software? I only find others they aren't as good as yours.
dip it in, and get it out, and then u get all the high speed compression gremlins.
hardware encoding is good at handling digitally recorded video sources. but when it comes to the analog sources with film grains... software encoding wins
what is the difference between Hardware and Software encoding.. and what they are named in Handbrake? am new to the program..
Never ever use MP3 for anything. It's been outdated since the late 90s. AAC has the same compatibility and much higher quality. For extreme size savings OPUS is a great choice too
Can anyone tell me what is the best way to reduce the weight of my video before uploading it to Instagram with HandBrake? Thanks!
Anybody suggesting that anybody should go over medium is crazy.
im a new subscriber
can u make other one for online study ,I use obs at 10fps ,and have a lot
Yea I initially wanted to also cover different types of videos like online lectures, teams recording etc but there were too many permutations and edge cases to cover.
You divide by 1024 not 1000 since you are in binary: 2^10=1024
So its closer to 0.73 GBytes for a minute.
CPU still give much smaller size and better quality
"Set audio coding to MP3"
I heard enough...
You lost me at Nvidia
😂
Hey,
great Video! really in depth an informative! bravo.
I have one question. Are your quality recommendations resoulution dependent? I want to convert 1080p and 2160p material with nvenc.
Thank you in advance! @Anitomicals
Using the quality slider is resolution independent. Files will come out with approximately the same quality with the same slider value regardless of resolution. (but of course it will end up at higher file sizes at higher resolutions)
To be resolution dependent, there is an option below the slider called AVG BITRATE (KBPS). input a number, usually from 80,000 for 4k and 20,000 for 1080p. it will hard cap your final file size to a desired size and hence will affect the output quality with different resolutions if you change the number. I picked the slider number exactly for this reason, I dont want people to fuss with different sizes at different resolutions.
@@Anitomicals thank you a lot! This clears everything up for me. This logic makes sense, and I am happy that they implement it like that.
I had this thinking because of the info text in handbrake when hovering over the quality slider. But it says it’s for h264 not h265. thats why I was asking in the first place to be sure about h265 :)
Handbrake is crap - learn the command lines for these tools and the advanced options and you can squeeze a lot better quality out
Could you tell me about that, please? 🤔