Video Artifact editor for V-DSLR and consumer cameras

Andrew Revvo Jan 4, 2017

  1. Video Artifact is a new standalone video editor for digital film restoration and footage quality enhancement. Free and paid editions.

    http://www.videoartifact.com/va/

    This program helps you to prepare footages from consumer or professional video cameras, digicams and V-DSLR.

    [​IMG][​IMG]

    It can remove MJPEG, MPEG, H.264, H.265 compression artifacts and digital noise, improves sharpness and color resolution up to 16 bit 4:4:4.

    Some restoration features:
    • Script-based to work with unique cases. You can use many filters in any combination.
    • 50+ specialized restoration filters work in Y, U, V separately in 8/16 bit.
    • Possibility to use dozens of different type denoisers for different noise type.
    • Accurate denoising that keeps micro details with no "plastic" effect.
    • Removing of sharping halos.
    • Linear color high quality spline-based resize to resize up to 4K. Resize halos can be removed.
    • Precise halo-free radius-dependent sharping.
    • Film grain emulation and 3D LUT grading.
    • 4:2:0 8 bit to 4:2:2/4:4:4 16 bit chroma reconstruction.
    [​IMG][​IMG]

    It can be used in a post-production workflow if you use V-DSLR or consumer cameras. Currently 50+ camera models supported and hundreds of compatible models. Some special features:
    • Batch processing without GPU. Very fast: 5 fps for typical V-DSLR AVCHD source and 2 fps for 4K.
    • Automatic joining of MTS, MOV and MP4 spans and renaming to yyyymmdd_hhmmss.mov format.
    • Non-standard 16-255 to TV 16-235 levels conversion.
    • ITU-R BT.601 ↔ BT.709 floating point conversion.
    • NLE and DaVinci Resolve compatible workflow.
    • Mastering to webm VP9/Opus or mp4 H.264/AAC 1080p/720p/360p/240p.
    Examples with various restoration methods. You can download low compressed videos to see actual quality.

    More examples:

    [​IMG][​IMG]
    [​IMG][​IMG]
     
  2. Is this a commercial product built upon the works of people who who provided their filters under an open software license condition?
     
  3. It is open source software, but with commercial pricing. Source code is included. It uses open source AviSynth filters, but most filtering code is by myself. 3 years in development and optimization. For example - my own 16-bit YUV core is used. Very aggressive code optimization to work with Full HD and 4K. Most speed-critical code is pure C++. Custom algorithms.
     
  4. One would assume it is all C++ (or C) except of course for very critical code which should be in assembly.
     
  5. Cary, yes, of course, assembly is used in low-level filters. Middle-level filters like sharping are C++. Some code is always C++ because the bottleneck is not CPU, but RAM access or multithreading caching. The optimization is a key to get 5 fps processing speed instead of 0.5. The history begins when I tried AviSynth in my production for commercial films using V-DSLR some years ago. Now I use it in all my projects. And it is now rather optimized.
     
  6. You might want to take a look at Vapoursynth since AviSynth is limited to 32bit processing.
     
  7. When you make a Mac OS version, tell me and I'll buy it the same day.
     
  8. I`m same with Marc. Make it run on macOS and I`ll buy it.

    During 2016 a have a lot of pain when match DJI drone footage to Sony RAW, and it seams that this kind of software could help with it. But my studio is mac based, and no way we move to windows in next couple of years. VMWare will kill performance, so not a helper here.
     
  9. Marc, Val
    unfortunately it is Windows-only and there is no way to port it to Mac without complete redesigning of all components. But it requires no GPU card and can run on cheap computers.

    I have interest to port it to GPU processing but not in near future, because it requires many months of low-level development and investitions. I am just an indie developer, so have no resources to do it.
     
  10. Andrew, I`m not blame you for made your software win only. I`m sure you have lots of reasons to do so. But in it`s current state it will be useless for me(((

    PS just as a mad idea - maybe it will be not a lot of work to port it using wine (it`s a way to run windows software on a mac or linux)? I know that some software vendors go by this route and it works
     
  11. Andrew while this is great and it looks like you put a lot of effort into it , i,m not sure if its at the correct price point, using neat video OFX which i would imagine quite a few people own on this forum have this , and it sits directly inside your app or a combination of using inbuilt noise reduction and LAB nodes and some other techniques you can get most of the results above

    if you jump up a step you be using the DVO tools for this kind of work or something like dark energy

    so there is no real compelling reason to buy this as, its the same price as the neat video package for OFX but nearly double the price for editors who can get the same neat engine in there edit software for $99

    i,m also intrigued as how you manage to go from 4.2.0 colour sampling up to 4.4.4 during a upscale or even just during the conversion ...the jury is still out on weather its true 4.4.4 on a downscale from 4k , and that's with all the extra info from the downscale

    i can save a 4.2.0 GH4 footage out as a pro res 4.2.2 but the underlying sub sampling pattern would be the same ...so how does that work then
     
  12. gavin,
    About 4:4:4: it reconstructs missing chroma pixels using special algorithm, so edges of 4:2:0 to 4:4:4 are clean. Of course, just repack to another codec/format is not a solution.

    See this image: [​IMG]
    It is a chroma channel difference/before/after from Panasonic G7 camera.

    About NeatVideo: it is a good denoiser and can help in some artifact cases, but the program is not a denoiser. It is primary a tool to get video ready to montage from camera flash using many tasks: joining camera spans, repack to another codec/container, remove artifacts, denoise. It works in batch mode per all sources. Most software filters were designed to fix V-DSLR camera artifacts where noise is not the most visible issue. There are a lot of cameras with aliasing, edges coloring and very visible compression blocking artifacts. When I started it to develop, my first camera had horrible aliasing and only 18 mbps bitrate.

    GH4 is very good camera in this sense. It has very cinematic noise and no know important issues (as I played with different GH4 sources from real film shots).
     
    Last edited: Jan 4, 2017
  13. Do you have a white on this special algorithm ..it would be very interesting to read

    BTW i don,t own a GH4 it just a 4.2.0 camera
     
  14. gavin
    I use warp sharpening in U and V with thin edge mask from Y. This way cleans edge pixels. There are another ways to do it, for example using blurring or antialiasing, but my method works with very bad noised sources.
     

  15. warp sharpening if i remember correctly was a gimp scm that came quite a while ago it's is a very clever way to get rid of jagged edges its interesting to see there is now a filter constructed for video

    however i still don't see how it increases the sub sampling of an image by doing this
     
  16. gavin,
    warp sharping produces clean edges for aliased ones. The result is a smooth line after sawtooth. But image will look very strange if used for whole image, so edge masking is used to process only image edges and only in U and V. Visually the result looks like native 4:4:4 and can be used for green screening where it is important.

    Another 4:4:4 reconstruction part is a motion denoiser that summarizes many frames after motion compensation. After 4:2:0 image (if stored in 4:4:4) we have many image fragment averaging, so flat parts will be restored, too. Similar way is used in astronomical photo to restore stars from very noised sources. Motion images give us a way to get sub-pixel details. So the software also can restore complete destroyed details after video compression.

    Of course, if we shot two close colored pixels in 4:2:0, they will be complete blurred and there is no way to get them back. Only parts those can be restored are restored.
     
    Last edited: Jan 4, 2017
  17. Warp sharpening is a free filter that works both for Avisynth and Vapoursynth.
    Many filters already use GPU so I am not sure about the comment about the GPU.

    Exceptional claims like upscaling from 4:2:0 to 4:4:4 using a "special algorithm" I would take with a large grain of salt!

    I would strongly advice Andrew to be very clear about what he developed on his own and what software he includes which was made by others for free!
     
  18. Cary,
    thanks for tips. I am very pedantic for the legal issues. So all required GPL license terms are not affected. You should know that VideoArtifact is open source software.

    Also if you really sure you work with AviSynth and Vapoursynth better that you can get using specialized software that I developed, just use them yourself. Anyone can create anything using AviSynth or node-based editors or by compile them yourself. Anyone can even develop the same product I designed. You pay not for free software but for my experience and the result than can be reached using this experience. If you do not understand how it works just ask me.

    I have 20 year experience in computer graphic. Also most my processing code is pure original and very optimized in hundreds hours of development. For example, you cannot find my 4:2:0 to 4:4:4 code elsewhere. It is my complete invention. Of course I use open source components for open source software. Anyone can compile it and use. Anyone can create anything using video editor, for example. But why you not ask yourself you make any work or it is a third party result: developers who created for you operation system, browser, video and text editor or something.

    PS. Many hundreds of AviSynth users use my own 16/32-bit HDRCore filters. I am open source developer and have created many AviSynth stuff since 2011.
     
  19. what's happening is actually chroma smoothing its been around for ages used on green/blue screen for DV and other bad capture formats

    have a read

    http://www.lafcpug.org/Tutorials/basic_chroma_sample.html

    no extra info/detail in the signal is added its just a guess/estimate/redistribution

    i,m sure when we used to shoot HDCAM it was 3:1:1 and if you fired it in to a deck it did something similar to make it 4:2:2 but that like a million years ago :)

    i just think its bit of stretch saying 4:2:0 to 4:4:4 with out explaining it ......
     
  20. gavin,
    yes, estimation. And most in computer graphic is just a psychology of perception + combination of well knows simple ideas from 60th. And anything created in filtering software just combination of very simple filters. There is no any magic :)
     

Share This Page