This urgently needs a patch to fix the frame preview problem with image sequences. If I import a tiff sequence and set the default 25fps to the correct 50fps it shows the wrong frame in the original comparison preview.
Seams dead slow. Proteus upscaling (480p to FHD) takes 4,x seconds per frame vs. 0,11 sec/frame in the previous version (Apple M1 Pro). Sticking to 2.6.4. as this makes the app pretty much unusable. No realtime-preview either.
For example, if a 60-minute video is trimmed and upscaled by 5 minutes from 25 to 30 minutes, it will be output as a 35-minute video.
However, since the video is actually 5 minutes long, this will result in an error in some playback environments.
of course lol… the update will continue but on only one release this one the official. so maybe in one week, a new version will be out, with more bug fixes and more features added, and so on, like they do with PhotoAi actually. so maybe next week, we’ll have a 3.0.1.0 (no idea, …) etc… so i think you can remove the beta and early version, and stay on this release. and maybe keeping the 2.6.4 next to it, for stuff that doesn’t work well yet on this 3.0 if “needed”.
it’s very easy to know what the next updates will be focused on, it’s the “known issue” in the change log at the beginning of this thread.
Thats quick! On 2.6.4 I get around 0.20 sec/frame on my M1 Pro Mac when upscaling 720p to 1080p using Proteus. I would have thought a lower resolution source would take longer??
Hello, I have installed the official version 3.0.0 and I have problems and doubts.
The problem is that when I render a video, I don’t see the image side by side (or any other for that matter). It takes time to load the image and it crashes, the image remains frozen, although the software continues to work, but without seeing the image.
In previous versions there were several models, now only 4 models appear and I don’t see the option to download the others.
Another problem is that when I want to scale a video, the Enhance model does not work and a lock appears. I can only enable enhance when the video output is like the original. I should be able to scale the video and enhance it at the same time, it was always done like this, now it doesn’t work
Installation on Windows 11 without any problem.
Now I’m processing again 3 videos in parallel with exactly same settings (1280x720 → 1920x1080, 50fps, Proteus relative, H.265 Main10 NVIDIA), but processing time again got worse than with last beta release (has almost doubled), while CPU load increased from 11,5% to 15,5% for each ffmpeg, and GPU load @ 95% with GPU memory utilization increases also from 3/12 to 10/12 GB (RTX 3060).
With such an increase in resource consumption I’d have expected faster processing (output quality seem to be ok).
Yeah… they only included hardware encoded H.264. It cannot be encoded with a constant rate factor. It’s also very lossy at any rate you set it to. You’ll get better results if you output to something more lossless then use another program to transcode to H.264.
The random number that gets added to the output. (Original File Name_203272476_amq13) When creating batch files to convert image sequences back into movies, it’s impossible to know what the random number will be. Super annoying to have to rename the directories by hand, or manually input the folder name into the batch file. (Of course I wouldn’t have to output to images if the TVAI enabled ffmpeg was compiled with the libx265 encoder.)
When opening a sequence of images, it defaults to 25 FPS AND THEN LOCKS Chronos to the model chain if you try to change it to the correct FPS.
I get around all this by means of a, now huge, python script that calls the TVAI enabled ffmpeg, but if I had to use the GUI, it would be painful for these two reasons.