Being able to tune enhancements (and other) settings and then run a quick preview to see if they work well is a great feature.
However, it would be even more efficiently done if there was a mode available, that would simply run a continuous preview loop, say (nominally) 4 or 5 seconds long. While the loop was being rendered, it should be possible to make small adjustments that would be rendered into the loop immediately after the setting was changed. - The affected control would change color until that setting change was being rendered and viewable into the loop.
When the desired results are found the user could then click a check button and the settings would be copied into the configuration settings for doing a standard preview or Export operation.
This is not hugely different to what we have currently in the 3.x beta, but it would make the tuning process faster and more dynamic.
I support this proposal.
That is similar to what I’m doing at the moment: I select a short trim of for example 5 seconds which I than process with different setting (within up to 4minutes processing time). Unfortunatly VEAI crashes sometimes already with the 2nd export and I must restart the program.
The video enhancers are designed to work on a stream. The nature of MPEG compression is not so much frame-based as it is change-based. As such, it produces better results to generate a number of frames out of the video stream. From there, you can single-step frame-by-frame and examine them in detail. Bottom line: It is actually more reasonable to enhance and capture a short sequence.
Yes, it has that potential. And, when desirable results in the loop show up, the settings could be caught and transferred to a “standard” preview.
The only real implementation impediment would be changing enhancement algorithms in mid-stream. However, I believe that being able to make setting changes inside adjustment loop, within the context of one enhancement algorithm would be very handy, nonetheless.
i don’t have the chance to have a 3090 RTX, just a 1050ti,.
A 1 second preview on a 720p to 1080p video with a 1050ti take some times, between one and 2 seconds per frame. it’s worst for a 1080p->1080p.
after the 10th preview/testing of setting in proteus, you just want to do one thing : turning off the software.
On short video or pretty static one it’s useful, you have a result quickly.
when i’m doing a music video, i’m filming myself while playing the piano in up view (filming the keyboard and my hands, nothing else).
most of the time, the lighting is not great and i have some blur on the picture.
i use VEAI to remove the extra blur, make the picture more sharp and clean the extra noise.
as the camera is filming static, the picture never change, except my hands. so being able to have a quick one frame preview on one picture is enough.
that’s an example among others.
sometimes i just want to do small and not important video, just to have a quick HD result. spending one hour on preview is “boring”. I often go back on 2.6.4 because of the quick one frame preview, if i don’t need proteus auto.
And a 5-second preview going from SD to FHD using Proteus is faster, but after several retires playing with settings, it is slow enough so it begins to get tedious. Even so, I’m not certain how useful viewing a single frame in a typical video would be.
it’s working pretty well, when we’re used to it.
for example :
time taken to see the first image of a preview on a one second preview from 720p to 1080p on VEAI 3 with proteus (estimate turned on + auto off) (last beta) with a 1050ti : 13,48 seconds.
time to generate a 1 second preview : 29 seconds.
time taken to see 1st image with proteus (estimate turned on) with 2.6.4 : 8 seconds
time taken to generate a 1 second preview : 52 seconds (ahah).
video used : 720p , 30 fps.
so 2.6.4 = whole preview a lot slower, but 1st image to see ? much shorter
Going only by v3b’s behavior, I suspect it works in overlapping chunks. This means that there will be a lot of recursive reading necessary to read a frame and (especially in Proteus.) The only way for the process to “see” and enhance the image in a given frame is to read and analyze those that came before it and those following it. - I believe that is why an image isn’t put on the screen immediately. (-Please Note: This is only my assumption of what’s happening-)
