Once again Video AI has been causing me lot of lost time and work.
This time I am trying to upscale an old collection to 4K and do Frame Interpolation of 29.97 FPS vs the original 24.97 FPS. The results have been better then my expectation for such old film media.
Unfortunately Video AI can’t make it through the full 1 hour and 20 mins of video. I’m not sure if its the fault of Video AI or if the system is over heating, but 3 times I came back to my PC after letting it do the rendering for hours to find it had cashed and rebooted. Last one was just at around 65% video encoded when I last checked it. Feels like a punch to the gut having 20 hours of encoding time down the drain.
Anyway,
I’m doing Image Sequence instead now, this way if it crashes I got the files still and know which frame to start back at. But I have a few questions.
By doing the frame Frame Interpolation, will it make new frames for image sequence and if so will that mean the frame number I have in the output file would be incorrect to reference in a crash event to start from again for encoding ?
How do I put the image sequence back into video with sound ? I seen someone say ffmpeg but I’m not sure of how or what to do for that.
Hey Scott, let me run a few tests and Ill get back to you on this. Some other users might chime in with their experiences as well.
For the image sequence and audio sync, you can pull the image sequence into a video editing app and line it up and lay the audio underneath and sync it up. Then export out as a video file in your preferred codec and format.
Biggest thing for me right now is to do with the frames. I need to know if its creating new frames for the Frame Interpolation step as images and if that means if anything happens, I won’t know what part of the video for frame number to start back at.
Which is going to be an issue as I just check the image sequence folder, it’s already filled to 497GB of frames in tiff format and the video only at 9% rendered. So that 2TB drive isn’t gonna hold all them frames.
Yes, the app will create new frames to fill in and change that frame rate for playback. Depending on the conversion will determine where these new frames are inserted in the export process. Here is a test I ran with an image sequence with the frames numbered on input and converted from 23.976 to 29.97 fps. This is just the first 5 frames and the second frame is duplicated, and then it duplicates the original 6th frame and so on for the video.
Well I went to bed, letting the encoding do it’s thing. Woke up today and went to the computer just as it blue screened and crashed. I’m now backing up 1.5 TB of frames to the NAS. I’m hoping Video AI will pickup at the last frame it started and not need to start over again. Because its going to be hard to figure out which frame to start at that isn’t the duplicated or newly created one.
Well no luck, I can’t figure out what frame to start at to save the progress made before the crash.
I will have no choice but to start over again with another estimated time frame of 2 days. But I say 8 or 10 hours in, it will crash again with a reboot happening.
This time I am doing Rhea only so I can at least get the correct frames incase of a crash. If I can get the full video dump, I will put the frames back into Video AI and attempt to do a video file with Frame Interpolation set to 29.97 FPS.
Are you sure you want to give up so soon? Just pick a starting spot near where it ended. Some over lap is okay. If everything works out well, the images will get the same numbers. If not, batch renaming them to the right numbers is not too difficult.
If you want to be extra careful, you can start the next bunch of images in a new folder.
I couldn’t figure out the area even for over lap as the 1 hour and 20 min video was way off count for the new frames. Also when opening Video AI, it started to process the project on it’s own again from the beginning, over writing.
What it boils down to is time and cost. I know my projects take a long time (days) to render and encode. So when they fail / crash after 10 or more hours, resetting everything back to square one again. It’s very frustrating and it adds to my power bill by a big jump. Last time I did almost 2 hours worth of video, took 3 days to process, render, and encode. Jump my power bill up by $15. Make things worst is, I hardly get time these days to do anything on computer anymore with work and life the way it is. It just too bad I can’t set the settings. Click export, and it all just render and encode the new file without issues.
Doing the image sequence would of failed anyway if the system didn’t crash as the hard drive that I selected would of ran out of space.
This time I selected an 8TB drive on my NAS, it just finished exporting the images, total size of Image Sequence folder is now 3.66TB for 115,900 frames rendered to 4K. Now I am waiting for Video AI to load the images back in so I can try to export them as an MKV video file with Frame Interpolation set to 29.97 FPS, I just pray it will let me and that the file size will be in the GB’s not TB’s. So far it been 20 mins and the program is locked up trying to load the images in.
If I do manage to get this far, any suggestions on adding grain for a 4K video. I kind of think the video needs some grain but what is the normal value for best results in 4K ?