Automatic multithreading exports would be cool. Maybe a checkbox, and then it shows each processing in the output section, but if one fails the whole thing fails so any bugs causing exports to fail should be fixed… I guess it is possible to do this very manually by triming videos into sections then rejoining them after processing…
When finally will TOPAZ fix preview jumping around? It’s awfull!
8 Likes
I am often not sure when the procedure is complete. The green check mark seems to be too early. When I quit the application at that point I end up with a corrupted, none working file.
Being I have posted on every single release thread multiple times since 3.x, including my build and stats… I assumed it was extraneous info.
AMD 5800x, 32gig ddr 4000, Nvidia RTX 3080, NVME gen 4 drives.
Compared to what version? Yes.
It is not slower or faster, this is just benchmarks for others so others can gauge their speeds as well.
My point is to post my speeds and issues per version. Your title does not seem to say you work for Topaz, so I am a little confused why you feel entitled to tell others our feedback is useless? Actually, looking at everything you typed, I am a little confused why you responded at all, all you provided was useless criticism and furthered a hostile forum?
First of all thank you for your work. With my MBP - M2, VAI 3.1.5 works.
On my Mac Pro (2019),
3,3 GHz 12-Core Intel Xeon W
AMD Radeon Pro W5700X 16 GB
80 GB 2933 MHz DDR4
MacOS 13.2.1 (22D68)
VAI 3.1.5 still crashes the computer after a few minutes - tryed several models
2 Likes
Hey, I think a understand the point of david.123, even if i don’t agree with the way he acted.
he’s there since a long time, was part of the beta testing of veai , report bugs , try etc, but like some of us here, he became frustrated by this forum. the forum has been chaotic between the release of 3.x and today, and even more before, not even speaking about people who are here without having a licence (i’m not speaking about trial users). the amount of these people has increased a lot and come here to ask “free support” on the forum (that’s why i think there is now a "video Ai Owner tag on each avatar). I don’t agree with what he told you or did, but i can see it on some other members (including me), not even speaking about the ones who already left this forum. you’re here since recently, you’re not concerned , so don’t take this for yourself. it’s more related to some users, past or present, and the big mess that this forum can be sometimes.
all the best 
!
1 Like
To direct people to the bug report and ideas threads, I think Nipun’s first post should include a link to those as well.
Write bug reports and ideas threads link even under “Video Roadmap Update”.
2 Likes
Ok, I worked out my issue, which was at my end, but it looks like these newer version of Topaz could do with an additional error catching mechanism.
I found a missing image from the sequence export. I am hoping it is only one, its hard to tell from the 160k images. Upon import, it reads the entire image sequence, previews, behaves normally but prematurely exits the full sequence with no error or warning.
I am assuming this is a change at the backed somewhere. In previous versions, if a single image is missing it ignores it and keeps going - basically losing that frame. In a sense, crashing out is better - at least you pick up something is wrong - but i can see clear errors in the logs to Topaz can see whats happening - just isn’t communicating it to the user, hence why that took quite a while for me to work out what was happening.
Just as some feedback for future amendments to the image sequence handling. On teh positive side, a direct comparison to V2 now has Proteus at around 50% reduction in time to process, hence why I was trying to test output again, which is really good.
Note we are a long way into V3 now, 8 months actually since the first Betas, and there is still no ability to handle frame numbers in this version, meaning complete processing inside V3 still isn’t possible. The speed improvements are fantastic, but if I need to cut up the video into 37 segments and reassemble them, I still have to revert to V2 as V3 can’t do it still.
2 Likes
It would be useful if, after downloading a model file, the date of that file was set to the date the model was created.
Another option is to add a suffix with the technical model number.
almost same Mac, except 16 core and 96gb ram, VAI 3.1.5 still crasher the whole computer, its been unusable for a couple of versions…
Hi - can you please send me your logs from Help > Logging > Gather logs for support? Thank you!
Hi - if you could also send me your logs from Help > Logging > Gather logs for support, It’d be very much appreciated. Thank you!
My first attempts to remove double frames was not successful. I also wonder if it only removes exactly two identical frames sourrounded by changing frames or if it removes doubled frames in a long row too. None of those were removed properly on my end. 
1 Like
In version 3.1.5, the processor heats up 2 degrees less, 63-64. In version 3.1.4 it was 65-66 degrees. Transcoding has become a little faster. It used to be 2.7fps, it became 3.0fps. But, do you want to be sure that these improvements are not at the expense of quality?
I am using Apollo to convert video to 120fps. I have source files in 4K resolution. Apollo introduced the Replace Duplicate Frames feature. How to insert this setting. How can I know how many such frames are in my video, and whether they are there at all?
1 Like
The Apollo whitewashing effect has cleared up, and the “Remove duplicate frames” feature works fantastic!
Still no “Prevent OS sleep”.
From what I know, duplicate frames usually come from DVDs. Basically most DVDs are interlaced and the way to get full frames out of them requires creating duplicates. It gets more complicated and obscure every time I research it again.
4K sources should not have any duplicate frames unless they are there intentionally. You can check by going frame by frame in a motion scene. The most common duplicate frame pattern I have seen is one duplicate in every six frames.
1 Like
Friends, I just noticed that when upscaling a DVD at 4k60fps, Chronos Fast makes fewer mistakes than Apollo. But when I convert 4K videos to 60 or 120 fps, Apollo does a better job.
If someone knows the answer to this, write if I’m right! And another question. Is there a detailed instruction somewhere on how different AI models work? I can not understand in what cases which model is better to apply.
Just had a crash on the encode I was doing:
2023-02-15 12-47-06 Thread: 41012 Debug Updating video info {"sar":1,"framerate":23.97602695365636,"startNumber":1,"frames":193056,"progress":68,"status":2,"frame":133045,"procStatus":{"status":2,"eta":29884.2311831439,"fps":2.007115906754496,"message":" Model Loaded","pass":1,"error":"","progress":68,"frame":133045,"priority":3,"requestPos":3,"processorIndex":0}}
2023-02-15 12-47-06 Thread: 41012 Info OUT: 5 frame= 133045 fps= 1.95369
2023-02-15 12-47-08 Thread: 41012 Debug Updating video info {"sar":1,"framerate":23.97602695365636,"startNumber":1,"frames":193056,"progress":68,"status":2,"frame":133049,"procStatus":{"status":2,"eta":29793.122487062028,"fps":2.013122555501145,"message":" Model Loaded","pass":1,"error":"","progress":68,"frame":133049,"priority":3,"requestPos":3,"processorIndex":0}}
2023-02-15 12-47-08 Thread: 41012 Info OUT: 5 frame= 133049 fps= 2.06707
2023-02-15 12-47-20 Thread: 41012 Debug passProcDone 0 5 5 -1073741819 1
2023-02-15 12-47-20 Thread: 41012 Debug Updating video info {"sar":1,"framerate":23.97602695365636,"startNumber":1,"frames":193056,"progress":68,"status":0,"frame":133049,"procStatus":{"status":0,"eta":29793.122487062028,"fps":2.013122555501145,"message":"Unknown error","pass":1,"error":"Last FFmpeg messages:\n\nUnable to parse option value \"0\" as video rate\nUnable to parse option value \"0\" as video rate","progress":68,"frame":133049,"priority":3,"requestPos":3,"processorIndex":-1}}
2023-02-15 12-47-20 Thread: 41012 Info Removing video "xxxxx_chf3_prob3_temp.mp4" 4
2023-02-15 12-47-20 Thread: 41012 Info Close and delete path "xxxxxx_chf3_prob3_temp.mp4"
2023-02-15 12-47-20 Thread: 41012 Info EXITED: 5 -1073741819 1
2023-02-15 12-47-20 Thread: 41012 Info ~TProcess(): destroyed
2023-02-15 14-53-02 Thread: 41012 Info Opening log folder: xxxxx\AppData\Roaming\Topaz Labs LLC\Topaz Video AI\logs\
2023-02-15 14-54-20 Thread: 31280 Debug Open Called "" 0x7a30
2023-02-15 14-54-20 Thread: 41012 Info Removing video "xxxxxx.mp4" 3
That was upscaling a 1080p to 4k + proteus on manual settings on my AMD 5800x, 32gig(ddr4000), 3080.
I would like someone to explain replace duplicate frames option and when I would use this.
Also what does the sensitivity effect?
No manual or detailed description of anything for this program.
New version is working great.
If you do a Google search on what duplicate frames in a video looks like, you will then understand why video playback can look ‘jerky’ (not smooth) with duplicate frames. So, replacing them with new unique ‘in-between’ frames that are generated by the AI should smooth out the motion of the video. In addition, if you increase the frame rate of the video (for example, from 30 fps to 60 fps), the result should also greatly benefit by replacing duplicate frames with new, unique frames.
You can check if your video has duplicate frames in any video player that allows you to advance the playback frame by frame. Choose a scene that has something moving for a couple of seconds, then you will be able to see duplicate frames, by advancing the video frame by frame. Usually, it is every fourth or fifth frame, although it can also be random.
Sometimes, the duplicate frames are not entirely identical, due to noise or something else, in which case you would use a higher sensitivity for TVAI to detect the duplicates. The drawback to that might be that it may select some unique frames as duplicates by mistake.
1 Like
