Video Artifact editor for V-DSLR and consumer cameras

Andrew Revvo Jan 4, 2017

  1. Ah, that's a shame. Maybe a solution for you would be to consider creating an OFX plug-in, which will work on a wide range of software on Mac, Windows, and Linux.
     
  2. This is very cool. Right now I'm using things like "notch filter" in nuke reformat nodes to deal with this stuff, but this could work much better for specific problems. Is there any possibility of some type of gizmo in nuke for this? Heres the common problems I use notch for and it's not applicable in about half the cases:
    1) expanding 8 bit to 16 bit color space (especially going from a 8k to 12k stitched 8 bit "image" to a crop at 2k prores 4444)
    2) camera operator screw ups, like moire or wierd artifacts like gate flares
    3) compression artificats (the nuke notch works fairly really well going from quadhd to 1080p, but I don't have anything that can deal with compression artifacts when I'm not reducing pixel count)

    I btw do compositing on windows (grading is in mac) with a z840 m6000, so having this run on the "compositing side" before it it sent to grading is where it would fit in my workflow.
     
  3. I have experimented with Boris DV Fixer, which can work to a point with crappy interlaced material. That plus some careful NR and sharpening helps, but there are limits in turning chicken manure into chicken soup.
     
    Patrick Faith likes this.
  4. I can suggest this workflow for users with Mac who still use consumer sources:

    Compile very cheap AMD 6 Core or i5 4 core computer: a cheap MB with USB3, a processor, good cooler, a cheap power block. An internal video card is ok, 8 GB ram, 1 SSD disk and 1-2 USB3 disks for processing and material exchange between computers. A very cheap Full HD monitor is sufficient, because it if for artifacts, not for grading. Windows Professional x64. This will give ~3-10 fps for Full HD sources and 1-5 fps for 4K.

    I personally use 2 computers, but I have hundred hours of sources from cheap consumer cameras. Any good camera for my work costs minimal 5x money than my software costs. I create education psychological films. And now I shot in 50p from 2 cameras + audio recorder.

    My workflow gives me simple transition: MTS files from 2 cameras → Video Artifact source prepare and restoration in batch mode → MOV → DaVinci mound and grade → DNxHD → Video Artifact mastering → consumer distribution webm VP9 in 1080/720/360/240p.

    Also I can extract WAV files, denoise them or recorder sourcs in an audio editor using best denoiser software and return resulted quality audio back to Video Artifact.

    In mount and grade phase I have clean sources with small film grain amount. No any filtering is required. So I can grade and mount in real-time for 50p 10-bit sources.

    I tried to emulate some features on DaVinci Resolve and Fusion, but it is unreal slow and it is not ready for practical use on my side. I had 5 fps only for reading and write with very simple filters. Any external OFX will dramatically slowdown processing up to 1 fps where Video Artifact gives 20 fps. I do not understand how to use these OFX stuff for practical needs without a great multi-computer farm with very expensive setup. Any GPU to CPU and back transition eats many processing time. So I understand why some restoration software have no prices.
     
    Last edited: Jan 6, 2017
  5. Right off the bat, I'd suggest you avoid shooting on cameras that record M2TS files. To me, M2TS is a delivery format for disc, not a format for post. Find a camera that will create reasonable files that easily work in your editing and color-correction workflow. As one example, the Blackmagic Pocket Camera will shoot 10-bit ProRes 444 right out of the box. When you start off with a highly-compressed 8-bit format, it's downhill no matter what you do to try to save it.

    The inexpensive Canon/Nikon DSLRs are not great, but they do at least have reasonable color and dynamic range and are workable to a point. They record to 8-bit H.264, but I'd say this is a better choice than MTS.
     
  6. Mark. I need 6 hour recording without big pauses with very small budget. Shooting ProRes 444 will give me only small fragments and tons of additional equipment to keep these fragments and make pauses. And setup price will be x50. It is not a cinema studio. I record realtime actions. Many users shot a similar way, even with cheap cameras, not V-DSLR.
     
  7. Andrew,

    Is there a tutorial on how to use this software? I'm just not familiar with open source programs.

    Thanks.
     
  8. Eugene ,
    yes. Currently there is one tutorial on YouTube in Russian voice + English subtitles:



    Also, right now I am creating two tutorials on how to use denoisers in free and paid editions. These tutorials will be published tomorrow.
     
  9. Video Artifact is very stable now. The current version is 1.6.0.
    Using VA I have already processed around 50 hours of video from different sources. 2 new education films were created using these processed sources.

    Some very useful Video Artifact Premium key features users usually do not know:

    Mastering for YouTube, mp4 and for webm. Latest two weeks I have improved these features. VA allows to encode a final master video from NLE or DaVinci Resolve for different purposes:
    • High quality YouTube export, compatible with 10-bit video. Usually 8-bit video is dithered and some values between bytes should be encoded with dithering patterns. For YouTube compressor this means additional bitrate for these flat areas. So we have compression artifacts even for static video. I had experiments with encode dithering. And now VA has the best possible quality encode algorithm that is compatible with screen-recorded content or reclaim video with large flat areas.
    • Consumer video encode for mp4. VA uses external x264 and Apple Quick Time AAC encoders. Currently we can create these video versions: 240p, 360p, 480p, 720p and 1080p, 8 or 10 bit. Correct color matrix is used: bt.709 or bt.601. Compression settings are tuned for a compromise between quality, encode speed and old device compatible. There are different encode scripts for screen content and movie with very different encode settings.
    • Internet video encode for webm. VA uses a reference Google vpxenc VP9 encoder and Opus. 240p, 360p, 480p, 720p and 1080p are supported. Compression settings were tuned for screen content and movie. Now I have best possible quality encoding per different resolutions. Every resolution encoding settings were optimized by hand. Also it is very fast. Default settings allows you to encode 1080p VP9 video on 4 GHz i7 with 2-4 fps.

    24-bit audio support. VA can extract audio track to edit in external audio editor. After editing (denoise, replace audio from external audio recorder) we can import these audio files back to get 10-bit video with 24-bit audio to edit in DaVinci Resolve or other NLE.
     
    Last edited: Feb 5, 2017

Share This Page