Hello,
Could you send us a note at help@topazlabs.com with the full log files from the app?
To gather logs, please select Help > Logging and make sure that there is a check mark next to File Logging. Next, recreate the issue and then return to the Help menu and select > Get Logs for Support and attach the “logsForSupport” file to your reply.
Alright.
But whatever was changed affects the quality of the results. Now everything looks much softer, blurrier than before.
Could you send me a direct message with your log files to see if the model scale is affected by this change?
Sorry, I’ve gone back to the previous version already.
I’ve confirmed that this will be fixed in next week’s update! (Friday for beta testers)
In Shakespearean Englsih:
It’s the automatic values yond art setteth f’r the ‘slid’rs’ from frameth to frameth
Here is a comparison.
Same settings:
3.5.0
3.5.1
Also, the processing time nearly triples.
Is this a 480p to 1080p conversion?
This latest release seems to have done something to Apollo. On my machine, it’s only processing the first couple of seconds of the video I’m slowing down. It still produces a full-length video, but it’s simply a static image after 2-3 seconds of output. Same for Apollo Fast.
576p, this is a PAL-DVD
Ok thanks, I just ran a quick test and found that setting the Second Enhancement filter’s Intermediate Resolution to 1x should produce results equal to 3.5.0:
Regarding replacing duplicate frames and sensitivity slider. Is there a program or media info type reader that tells you how many duplicate frames you have?
Interlaced 720x480 DV to 1080/60… 13 hour render… video hangs after 10 seconds.
Not impressed.
Yes, I think so too. If we take the case of neat video which uses 11 images, (5 before and 5 after) it is better because it is for temporal denoising, it lets in fewer artifacts, and normally to the detriment of a loss of details but strangely I find that neatvideo is perfect with the radius at 5 because I never noticed a loss of details, but always better quality, not like the denoise or degrain filters of Hybrid, when we increase the temporal, we lose details, qtgmc also by the way, “final temporal smoothing”, that’s why I always leave it at zero, because from the moment we define 1 or more, these are details which are lost. But anyway, I digress… For auto iris prediction with fewer images it’s the opposite, better because it’s to analyze the entire image with sharpness, dehalo, compression, and if the average is done with fewer images, I think it is more precise.
Thank you. I’ll try that.
This works! Such a little change, bit difference.
Thank you. 
Tried it with ARC A380 (latest drivers Intel 31.0.101.4826 Beta and Win11 22H2 updates), and the “Iris-v2 is supported on Intel ARC.” works indeed on 2 different machines with different CPUs (AMD/Intel).
Tried on an ARC A770 16GB on a 3rd machine (also the same drivers/updates) and couldn’t make it work in any case, and codecs combinations I’ve tried (h264/5/AV1 on both Intel ARC or NVIDIA)
All three machines have RTX cards as the main ones and ARC are installed only to the AV1 & Quicksync capabilities. Very strange that it doesn’t work only on the A770 GPU.
Intel 31.0.101.4826 Beta and 31.0.101.4824 WHQL were used for ARC A770, but all attempts produced either error @ Topaz Video or multiple blue screens with VIDEO_MEMORY_ALLOCATION (if I can recall correctly)
Coming from 3.4.1 to 3.5.1 I cannot upscale 1080p to 4k anymore.
I never had any issues with this but with 3.5.1 every video I tried resulted in an “out of memory” error within the first few seconds.
I updated my nVidia drivers after the first errors to the latest version but this didn’t change anything.
Windows 10, nVidia RTX 2060 Super, 16 GB RAM, Intel i5-7500
Hello,
Could you send us a note at help@topazlabs.com with the full log files from the app?
To gather logs, please select Help > Logging and make sure that there is a check mark next to File Logging. Next, recreate the issue and then return to the Help menu and select > Get Logs for Support and attach the “logsForSupport” file to your reply.
I don’t think there is. There is no way that I’m aware of to automatically and reliably tell the difference between a few frames in a video where movement briefly and correctly stops, and unwanted duplicate frames. Same thing applies with ‘missing’ frames.
It should be possible in sections where there is clearly continuous movement that suddenly stops for a frame or two (or jumps at a missing frame) but where would developers draw the line? That’s the problem.
What I’d like to see is a way to mark the location of dupiicates (or missings) when identifying them manually - and for TVAI to then interpolate new frames or remove duplicates according to your markers. That would be really useful on short video clips. Some degree of automation might be possible but in the end, it should be overrideable with manual markers.
EDIT: This kind of thing would be a good fit for TVAI, which aims to help us restore and improve video that’s often in poor condition.
