The Mac version is showing ‘Error’ when trying to encode using a number of either existing presets or user made presets. Also the benchmarks mostly now return errors
Topaz Video AI v3.2.6
System Information
OS: Mac v13.0301
CPU: Apple M1 Max 64 GB
GPU: Apple M1 Max 48 GB
Processing Settings
device: 0 vram: 1 instances: 1
Input Resolution: 768x576
Benchmark Results
Artemis 1X: ERR fps 2X: ERR fps 4X: 04.60 fps
Proteus 1X: ERR fps 2X: ERR fps 4X: ERR fps
Gaia 1X: ERR fps 2X: ERR fps 4X: ERR fps
4X Slowmo Apollo: 24.30 fps APFast: 88.90 fps Chronos: ERR fps CHFast: 10.24 fps
Some videos and image sequences are being exported with 2 frames less even though I’m only doing upsampling (no motion blur or time interpolation). I confirmed the fps setting matches the original and it happens if I export either in a video format or image sequence. Original video/image sequence is 144 frames and the export is 142.
Copying the ffmpeg command doesn’t seem to be working this version.
I’ve tried the copy button and highlighting and cmd-c on both export and preview commands from both the menu and the ellipses on the preview list.
Yep, here too, Macbook with M2 Max, latest update rendered Video AI completely useless since every encoding and preview results in the red X. This doesn’t look very good…
Having spoken to Ida @ Topaz support about this, I had to do a full uninstall of v3.2.6 (I use AppCleaner which will delete plists etc) and then I re-downloaded the app as there was a fix added for the error issue. Upon re-installing and re-creating the profile I normally use, I now have it up and running and batch encoding a number of files without the error issue, so if you do the same it should sort out that problem.
We have upgraded from Video AI version 3.2.3 to 3.2.6 and we have found that when rescaling an HD EXR sequence to 4K EXR, the new sequence always omits the last 2 files after the job is finished. This is something that did not happen in version 3.2.3. Regards.
Seeing a huge CPU usage spike in ffmpeg.exe in task manager when previewing, this was not in 3.2.4 (i didnt test 3.2.5 so it could have been there as well). Is this part of the new color correction? I didnt have this problem before and its bogging my system down.
… which reminds me, there’s a recent Ideas thread for an option to fill in missing frames, in which I’ve commented there and on a previous occasion.
Well, scene change detection would of course require TVAI to identify sudden changes in frame content - which is exactly what would be needed to pick up ‘missing frames’ in a dynamic scene, but with different settings. Big scene changes for a real scene change, very small changes for a missing frame which, once identified, could be interpolated.
TVAI would have to continuously monitor the rate of content change from frame to frame and be able to pick it up if that rate suddenly increased then immediately returned to its previous rate. So it might be a slow process and usable only for short clips but still, we’d have the option.
The frame preceding a sudden jump could then be marked/noted by TVAI and the next (missing) frame interpolated either at the same time or in a subsequent run. It wouldnt work on static scenes with no movement but that shouldn’t be an issue as missing frames then are not usually noticeable anyway: movement is the problem.
I believe this could be done, with controls to adjust the degree of sensitivity as with removing dups, and it would be an important enhancement for many people who are sensitive to missing frames as well as dups - both being especially annoying when using Apollo or Chronos. I linked to a couple of examples of scripts intended to do just this, in the ideas thread.
So please, Topaz, let’s have both! Scene change detection first, and then missing frame detection and interpolation.
Hi, I’ve stuck with version 2.6.4 for a long time on an AMD Ryzen 7 5800H with a GeForce RTX 3060 and 8GB of ram, would 3.2.6 be an improvement worth investing in?
It’s a wait and see thing for me. I read the forum and continue to see problems others are experiencing with that latest version, which leads me to believe the software should still be in the beta testing phase, and not a commercial release. I won’t pay good money to be a beta tester. Once I see measurable improvement, I’ll consider upgrading, but not until then. Through this process, I’ve learned that AI is nowhere close to taking SD video and upscaling to content, which rivals material originally mastered in 4K or higher. I got caught-up in the AI hype. The few television shows that I’ve upscaled are basically enlarged versions of the same thing, and nothing like I imagined when I first bought the software. Granted, it looks a little better, but nothing like content originally mastered in 4K.
That was the reason for my question. I’ve still been seeing people commenting on issues, and no-one saying that this latest version or any other version is a significant improvement in either quality or speed.
I’ve been upscaling old TV shows and films, and while there is an improvement, the best I can do is a 2x upscale. The software is not yet at the level I expected when I first got swept up by the AI video upscaling hype.
3.2.x is 2x faster than 2.6.4 and have few new interesting improvement but if You must pay to have new version I think it is better to wait for version with face and text model improvement.
You see with your eyes… the program can’t do that, everything has to be estimated about what it is… should it be enhanced or attenuated. Is it noise or detail.
What Topaz video AI can do is amazing.
But I have a few things I would like to see:
1.) Please keep the zoom and position x,y for all rendered clips. It makes no sense to save the zoom for each clip, because if I want to compare different settings it sucks if x,y and zoom are always different.
2.) Please a ‘Save project’ and ‘Load project’ option.
3.) A dream would be a trim batch list, where several trims of a clip are processed.
When exporting TIF files program deletes all completed frames if there is no space left. I have preferred images to have something to use WHEN the program collapses or in a situation like this. I could use those TIFs to build a video.