Easier cropping / duplicate frame removal

Two ideas:

  1. An interface convenience related to cropping:

When highlighting multiple videos, many settings can be applied to all of them at the same time (framerate, model, resolution, etc), which is great! However, any crop-setting will NOT be applied to all the videos, just one at a time. Even if every video requires individual adjustment of the size of the crop frame, it would still be more convenient to select (for example) Crop → 16:9 and apply those parameters to all the highlighted videos. Ideally, the full 16:9 crop-frame would then outline every video, allowing for more efficient individual adjustments. The way it is now, one must repeat the process of first clicking “Crop” and then choosing the aspect ratio for every single file.

  1. Duplicate frame removal:

In my experience, the feature to remove duplicate frames has some room for improvement. It seems to render blurry “in between”-frames rather than actually removing duplicates. I think it would be better if the program would actually weed out duplicate frames and then “stretch” the remaining ones to maintain the video’s original runtime. This would of course require timestamp modification of the remaining frames, possibly resulting in an unusual framerate, but that could later be modified with the Chronos or Apollo models. Strictly speaking, such a method would cause slight A/V sync issues, but the frametime is probably too short for that to be noticeable.

Based on testing and comments from the developers, duplicate frame removal seems to work like this:

  1. The system detects duplicate frames.
  2. The system then removes them and replaces them with a frame generated by the AI model you selected (E.G. Apollo or Chronos).

So it seems it’s already doing generally what you want.

The problem is that the currently available method inserts a blurry frame rather than removing the duplicate.

Using another program to manually weed them out from the source file in advance yields much better results, but is just too impractical to be feasible.

That also applies to missing frame generation - it can be done manually but you have to first render as an image sequence, then manually identify where each missing frame should be, then use TVAI to create it by interpolating from the 2 surrounding frames, then slot the interpolated frames into the image sequence in the right places and finally, render the ‘fixed’ file.

So it’s even slower than manual dup removal and yes, completely impractical for anything more than a very short clip. Plus, some videos contain both duplicate frames and missing frames and I’ve fixed that manually, with excellent results. For a 5 seconds’ clip! :laughing:

Even a semi-automatic method would be a good start on shortish clips (manually mark the location of each missing frame/pattern and/or dup, and let TVAI do the rest) so come on Topaz, we need you here!

Yeah, I just did another test run, trying different framerates, but the results were the same as before. The output videos all had series of very blurry frames in addition to a very strange “rhythm”. The output files were worse than the source material in every way.

The least bad option I’ve found is to use another editor prior to TVAI, which sometimes works perfectly (but it usually removes duplicates from just parts of the video).

All in all, TVAI is certainly a wonderful program! But if there is a way to get the current duplicate removal feature to work properly, I’ve yet to find it :slightly_smiling_face: