I use the free software VirtualDub. Just grab the first image and it will load all of them. After that you can encode or just save it to a video file. There are options to choose the exact framerate and you can add the audio layer if you have extracted it from the original file before.
Ich benutze die freeware VirtualDub dafür. Wähle das erstes Bild und es läd alle Bilder rein. Danach kannst Du es als Video neu encoden oder einfach erstmal speichern. Die Bilder pro Sekunde kannst Du präzise einstellen und auch eine Audiospur hinzufügen.
Looks like there could be an issue with handling framerates? Whenever I input a video file running at 24fps, the output file comes out at 29fps, in turn de syncing the audio from the video.
I have used VE AI from the very first non GUI beta version to the current one, and I’m delighted with all of the amazing improvements that have been made. As good as it is, there is still room for improvement, most of which should be fairly easy to implement -except the very last one
FUTURE FEATURES SUGGESTIONS
First of all, the Labels for the various Processing Modes are very poor. Good labels should be very clear and fully informative, rather than requiring a table of definitions to interpret them. Names should be FUNCTIONAL - as few abbreviations as possible. Granted in this case finding functional names is somewhat difficult, but they certainly could be better.
Second, although possible, it is very difficult to create multiple start and stop points for still frame extraction. There should be a RESET button to clear the next set points. Possibly a button for marking individual non sequential frames.
Third, there should be a GUI Command for turning a sequence of extracted still frames back into a video -rather than having this function available only if you understand the batch processing commands fully.
Fourth, and possibly more important than the prior points, Topaz Video Enhanced AI should be “translated” to work with the more powerful Radeon video cards which are often much less costly than those from nVidia.
Noticing some weird things relating to frame-rate outputs. Anyone else feel like its somewhat inconsistent? some scenes come out fine, and others stutter or seem just… inconsistent?
I experienced similar things when I re encode video files via the Smooth Video Project (SVP), but when I play it again and again, parts of the movie that stuttered before maybe don’t at the next attempt, so I assume in this case it has something to do with the video player.
I wish it was my video player, however my video file stutters wether its on my computers, or through a plex media player. It’s really rare that it has lower fps, but it’s definitely there.
But careful, it is does not working for solving the problem at crashing with a video output. It starts from the beggining and it overwrites automatically the last mp4 file.
I recommend using jpg or png output. If the program crash, its easy to continue starting from the next frame number after the last proccesed. And then make a video using another software. Like Virtualdub, Avidemux, Blender or FFmpeg command line. And remember to put the same fps of the original video.
Another thing i’ve noticed is what i’d call “jittery” frames? In the latest version videos have some scenes where a characters teeth might “jitter” back and forth etc. I really dont know how to explain this without a video, but anyone else see anything like this?
Thanks for the quick response. I think you’re right on the money. I’ll do some looking into how to get them detelecined , now that I think about it, this should stop what appears as “framerate lag” that I posted about before.
edit: Here’s a GIF I took of my source video with the ‘jittering’ on the character sitting down. though ive run it through programs to detelecine it, it still stays. honestly have no idea how to solve it.
Positives
The higher bit rates for MP4 output are welcomed. In previous releases, bit rates were sometimes lower than the original file. A work around was outputting the files to 4k and downscaling to 1080p. I felt this reduces the number of macroblocks.
Negatives
The audio is not syncing correctly. It doesn’t matter if I try to let the program sync the audio or if I mux the audio after the conversion is complete. It’s basically unusable. The audio is seconds out of sync and no manipulation in post is correcting the issue. I had to uninstall 1.5.1 and revert to 1.4.2, which does not have this issue but output bit rates are lower. I tried processing the same file with different models and it yielded the same out of sync results.
I’m not sure if other have noticed but I have noticed some distortion in the background of the image in the Theia models. The actors look fine but the background, in some cases, looks like its waving like a flag. It’s especially noticeable in scenes with lots of foliage. This is not present in Gaia model, which is the model I have defaulted to and the one I feel yields the best overall results. It looks like the distortion problem was corrected for 1.5.1 but the audio problem makes that release unusable.
Those look like frames with interlacing artifacts. Which means the source is telecined. During 3:2 pulldown, the DVD player will discard those frames. They’re there for playback on NTSC televisions that need an interlaced signal (i.e., old CRT televisions).
Change settings in the numbered order shown on the screenshots.
In the first image, after number 1 (change to AviSynth), click the button directly underneath of it (has “AVS” in it).
The previewer comes up. Seek for video frames where there are lots of motion (camera pans, hand movements, etc.). Step through the frames one at a time in the previewer. If you see three progressive frames (no artifacts) followed by two interlaced frames (artifacts) and so on, then the DVD is probably flagged as 29.97 fps and the correct progressive scan rate is 23.976 fps. To detelecine to 23.976 (remove the interlaced frames), use the settings as indicated on the first screenshot. Change settings in numbered order.
The second two screenshots show the encoding modes I recommend for creating an .mp4 file you can then import into Video Enhance AI.
The second screenshot is settings for CRF of 1. This is what I use and find it works just as good as lossless and saves a lot of hard disk space, if that’s a concern. (If you output from Video Enhance AI as .png’s that typically takes a lot of space!)
If you want to do lossless, use the settings in the third screenshot. Select settings in numbered order.
I process my video file to images. When I load them into VirtualDub I keep sure I make the video the exact same frame rate as the original file. I extract the original audio and add it via VirtualDub again. This worked well with an 1 h 40 M long movie recently!
