1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Windows Simple to use free video editor with GPU encoding

Discussion in 'Software' started by wyx087, 7 Feb 2021.

  1. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    Any recommendation for a very simple to use video editor with GPU encoding.

    My parents are digitising very old camcorder tapes. They are digitising it. They will edit the video using some sort of editor. Finally give me the portable HDD to encode it and put it into the family album (hosted on my NAS)

    First, the editor must work on their old Phantom II 6 core computer with integrated graphics. (When digitising, they use VLC to save it as AVI, even x264 MP4 is a bit too much!)
    Second, the editor need to be very easy to use, they want to trim a video and place it one after another into a single video.
    Third, the editor will need to have a project mode so I can open the project on my machine and apply all the edits.
    Finally, it needs to be able to export using NVEnc x265.


    I looked at OpenShot video editor, but just to encode a 3min video it's taking aaaaaagggeees using the CPU. They say v2.5 supports NVEnc but only on Linux. The interface and ease-of-use looks spot on though.

    DaVinci Resolve looks super complicated to use..... but does it do everything else?

    Anything else? I remember Windows Movie Maker, it's that level of editing, but with NVEnc support.
     
  2. Arboreal

    Arboreal Keeper of the Electric Currants

    Joined:
    21 Jan 2011
    Posts:
    4,150
    Likes Received:
    845
    My wife uses Filmora 9, it's free for starters and IIRC pay once you're happy with what it does to get un watermarked results from memory.
     
    wyx087 likes this.
  3. VipersGratitude

    VipersGratitude Multimodder

    Joined:
    4 Mar 2008
    Posts:
    3,450
    Likes Received:
    775
  4. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    Thank you, both.

    VideoProc doesn't seem to allow saving the edits. Does support x265 with GPU encoding.

    Filmora worked great in my quick test, ticks all the boxes. Plus they can use free version to edit on their machine and I only need to buy one license for export. Hopefully works okay on their ancient machine.
     
    Arboreal likes this.
  5. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    Just one thing to watch with NVENC: it's fast, but it's terribly inefficient. I did a quick H.264 test a while back on a teeny-tiny AVI, and while NVENC was four times faster (800FPS compared with 200FPS on the CPU) the resulting file was 45 percent larger at the same quality settings.
     
  6. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    How does that work? If I set bitrate to be 8Mbps, both output files should be the same size?
     
  7. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    I was using the default setting on both encoders, of "constant quality" - not a pre-set bitrate. I can try again with a pre-set bitrate...

    EDIT:
    Same file, two-pass encoding with turbo first pass and a 6000kbps target bitrate: NVENC averages 1,800 FPS on the second pass, finished file is 1.4GB; same settings but on the CPU gets me 315 FPS on the second pass with a finished file of 795.4MB.

    Even worse that time: NVENC's version is 76 percent bigger!
     
    Last edited: 8 Feb 2021
  8. veato

    veato I should be working

    Joined:
    15 Jan 2010
    Posts:
    766
    Likes Received:
    91
    wyx087 likes this.
  9. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    I'd thought bitrate always governs the size of video file?

    I ran 2 tests using my usual encoder, Hybrid, where I have always used NVEnc. Files out of it are similar size, targeting 5 Mbps bitrate.

    upload_2021-2-8_9-51-39.png

    Massive speed differences though, 15 FPS vs almost 200 FPS.

    CPU setting at default, changed encoding mode to average bitrate and used 5 Mbps
    upload_2021-2-8_9-53-18.png

    GPU setting at default, only changed bitrate settings
    upload_2021-2-8_9-54-17.png
     
  10. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
  11. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    I don't believe there's such a thing as a fixed bitrate in modern encoders; they're all variable bitrate, and take their own decisions about when to vary it and how much by.

    I'm not alone in finding NVENC great at speed and terrible at everything else: Handbrake's benchmarking found NVENC poor for quality and file size, this Reddit user saw a file encoded to 4.15GB on the CPU and 11GB on NVENC, this guy found NVENC averaged out at over double the bitrate compared to CPU encoding, this guy's recompressions were larger than the source files...
     
  12. jinq-sea

    jinq-sea 'write that down in your copy book' Super Moderator

    Joined:
    15 Oct 2012
    Posts:
    8,668
    Likes Received:
    616
    ^ This, in general. I used to prosecute the patent portfolio of a big player in this space.
     
  13. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    But there is a target bitrate where the encoder will try to hit with the overall average bitrate.

    I'd be interested in media info of your 2 comparison files, feels like the output is at vastly different bitrate.

    I did try Handbrake initially, and also found inefficiently large files from GPU encoding, that was using Maxwell GPU. Your links suggest problem still persist with Turing. I think the problem might be CUDA implementation? Hybrid and Shadowplay files are both reasonable sized and good enough quality. The Hybrid export I did earlier does seem ever so slightly softer in the in-focus details of my jumper compared to the CPU results. I can live with this very slight blurriness (only noticeable when paused and pixel peeping) for the much faster speed.


    Filmora x265 HEVC gives me an error when disabling hardware acceleration, so I can't compare output quality/size.
    Pinnacle Studio feels rather buggy, haven't managed a single successful export yet, It feels like it's geared more toward selling effects DLC.
     
  14. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    It was (naturally, 'cos bitrate is equal to file size divided by media length in seconds...) - but, as I say, they were both created with a target 6,000kbps bitrate (Handbrake's default.) NVENC was closer to the target, and CPU was significantly below - quite likely 'cos I was using a terrible quality low-resolution 91MB AVI from... 2005, apparently, as the source.
    Yup, I'm using Handbrake and have an RTX 2080 GPU.
    For me, I'm willing to take the hit and compress on the CPU in order to maximise quality and file size - but it's horses for courses. If I were streaming, I'd use NVENC 'cos who cares how big the output is so long as it's not a higher bitrate than my available upstream bandwidth?
     
  15. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    Wait, that would suggest CPU encoder isn't doing its job properly?

    There are 2 ways to look at this, I have set a bitrate, the encoder should hit that, or should bitrate be adjusted automatically based on input quality?

    If you feed an input video that is vastly better than target output, then the 2 encoder should output similar file size?
    The test video I used was from Fuji X-E2 1080p 35Mbps, down graded to 6Mbps, generally I find this good enough for mobile phone /tablet viewing.
     
  16. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    You could argue it either way - but as far as I can tell, based on others' experiences, it's NVENC that ain't pulling its weight.
    There's only one way to find out: let's give it something to really sink its teeth into!

    EDIT:
    "The requested URL was not found on this server." Marvellous, let's try something else... Nope, that whole website is busted. This is working - I'd have preferred something shorter, but whatever.

    EDIT EDIT:
    Right, CPU first: 6000kbps target bitrate, VFR, 2-pass encoding disabled, no other changes. 42 FPS average. Final file is 63.3MB, 5,315kbps - within shouting distance of the target, especially as the whole thing's only a minute and a half long.

    NVENC now: 6000kbps target bitrate, VFR, 2-pass encoding disabled, no other changes. 51 FPS(?), my CPU's pegged - that's not usual. Final file is 71MB, 5,986kpbs - closer to the target, but there's only about 600kbps between the two. There's something it doesn't like about that source file, though: the CPU's getting hammered and the GPU's encoder's only showing 25% utilisation. Bizarre.

    EDIT EDIT EDIT:
    Figured it might be 'cos I was scaling the file from 4k to 1080p, so had another shot while retaining the source file's resolution - so it should be a straight reencode job.

    CPU: 29.5 FPS, 63.1MB, 5,301kpbs - so still 600-700kbps below target.
    NVENC: 65.12 FPS, 71MB, 5,986kbps - so, again, closer to the target. CPU still ran hot, but at least the GPU encoder was showing 100 percent utilisation this time.

    The only remaining test would be with a longer file, the target bitrate being an average over a period of time - dunno if the CPU encoder would get closer to the target then, or if there's always a shortfall. Ain't got time right now, though - I'm supposed to be writin' as it is!
     
    Last edited: 8 Feb 2021
  17. wyx087

    wyx087 Homeworld 3 is happening!!

    Joined:
    15 Aug 2007
    Posts:
    11,157
    Likes Received:
    375
    That's what I call a trippy video. Anyway.... Good for quality testing.

    Hybrid NVEnc x265 did 75 FPS with 100% GPU utilisation, 45% CPU utilisation. It was set to 6 Mbps. I got 66 MB output file, overall bitrate 5865 kbps.
    Hybrid x265 using Ryzen 3600, it's only doing 4.7 FRS (?!)...... average bitrate 1 pass, same 6 Mbps bitrate setting. I got 63 MB output file, overall bitrate 5580 kbps. (no resizing on both)

    Is the bitrate always supposed to have a shortfall? I understand able to produce better quality with less bits is an advantage, but that sort of defeats purpose of setting a bitrate?


    Got to say, the CPU file does do a cleaner job than the GPU. But this is peeping at 2x size.
    (click to embiggen, left to right: original, GPU, CPU)
    [​IMG]
     
  18. Gareth Halfacree

    Gareth Halfacree WIIGII! Staff Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    14,627
    Likes Received:
    3,459
    That dun't sound right: I'm on a 2700X, so I've only got two more cores and an ickle bit more boost clock than you yet I'm getting an order of magnitude faster encode on the CPU.

    EDIT:
    D'oh, you're doing H.265 - I'm doing H.264. If I switch to H.265... Yeah, I'm down at 8.5 FPS CPU and 65 FPS GPU.
    The idea behind setting the bitrate on a modern encoder is to give it an average to target. During scenes where not much is happening, it could be a fraction of your target; during busy scenes, it could be significantly more (if you look at the GPU encoding settings you posted, there's a "maximum bitrate" of 9,000kbps as well as the target bitrate of 5,000kbps).

    That's what two-pass encoding is for, incidentally: it runs through quickly to take a gander at which bits might need a higher bitrate, then steals some from other scenes in order to boost quality without changing the overall filesize. (On commercial Blu-ray/DVD releases, they have someone whose job it is to do this bit manually, believe it or not.)

    Now what's supposed to happen is that over a long enough period the higher and lower bitrates are supposed to average out at your target - but as we've both seen on that test video, it's averaging out about 10-15 percent lower. That could be 'cos a 90 second video just ain't long enough - the only way to find out would be to encode something longer and see what happens.
    That's the biggun for me: although NVENC appears to be doing a "better" job of hitting the target bitrate (on this short video, anyway), the quality is actually worse (albeit only a bit) than the encoder that's getting away with a lower bitrate.

    It's easy for me to say I'll use the CPU encoder, though - I'm getting 30 FPS, not 5 FPS!
     
    Last edited: 8 Feb 2021
    wyx087 likes this.

Share This Page