Sometimes i have black frames in a video, and VEAI still upscales them and wastes resources, it would be great to see it generate black frames, or just use the same, the same happens when i upscale a presentation and there is noise, multiple frames of the same image with little to no artifacts are generated, it would be great to just use the same frame across the repeated frames, in order to save time and resources, scene detection in proteus v3 could help with multiple parameters if they can be applied to the timeline
maybe a more agressive temporal denoise?
Hello,
Introduce a function that allows you to analyze the scenes and cut the scenes on a time line like other software (like premiere pro and co, so upstream you have to introduce a timeline in TOPAZ).
This would introduce the possibility of making a manual adjustment (with proteus/chronos/Apollo) on specific scenes which would deserve more correction than others (noise for example or Chronos fast better than chronos and vice versa).
The workflow would be to make an application of a parameter with a preview and if the parameter does not work give the possibility of making a change of parameter only on the problematic scene.
With this function would avoid an application of the same parameters for all the footage, but on the contrary allows a full manual control scene by scene (not like relative to auto function).
It’s a shame to have to make cuts on other software like premiere pro, when doing it on TOPAZ would be better.
THANKS
FYI - If you are willing to use FFmpeg CLI, there’s a post that describes how to preprocess a file into scene segments with FFmpeg at:
https://www.reddit.com/r/ffmpeg/comments/ws3dwj/segmenting_on_scene_change_boundaries/
The example uses H.264 codec, the TS container and FFmpeg’s segment muxer. Alternatively, FFV1 codec with nut container could be used to create a lossless intermediary file. Individual segments could then be processed in TVAI and then concatenated/reconstructed back to a contiguous file using FFmpeg’s ffconcat from the index file.
Each scene-file is named in ordinal sequence.
Unfortunately, the process requires some scripting and FFmpeg CLI skills. FFmpeg’s scdet can accept a threshold parameter to determine how aggressive the scene change detection is applied. scdet does not use AI, it is a threshold-based algo.
The process should allow a user to process different scenes with different filter (bwdif, pullup, tvai etc) and is especially useful for hybrid / mixed-mode content.
This would be especially useful for when upscaling power point presentations and would drastically cut down on render times.
