A remote RDP connection from my Linux Desktop to a Windows 11 machine. The thing is, the older version works fine. With the newer version of VideoAI if I connect to the machine it simply crashes the application; it becomes unresponsive. I wonder if it is running under a different context.
I could set up VNC, and deal with it, but I figured since it worked before it should work now.
Not sure, it seems that I updated the latest Nvidia studio driver 527.56
Now the speed from 1080>4K is increased from 1.4~1.6fps to 1.6~2.0fps 
(proteus relative to auto)
*google translate
Using Mac Mini M1 computers ā the machines are silent and use little power (16GB RAM costs $899 or so).
Regarding the actual speed it says it will take to render a video file ā it is completely different then if you were to use a āstop-watchā to measure the real human world time.
I have learned that if you provide TVAI or VEAI with industry standard video, not in a weird format shell, not with odd audio tracks, with square pixels, with a recognized frame rate and in a resolution that is standardized ā that you donāt run into crazy issues and extra long delays. Just my observations. And I de-interlace my video prior to sending it for up conversion. It is faster and I feel there are better applications for de-interlacing.
TVAI or VEAI ā I donāt know what term to use (as they share the same version numbering - but are considered different programs). Iām still confused why TVAI did not start at version 1.0.0 and VEAI stopped (obviously) at version 2.6.4?
VEAI was the old Video Enhance AI. Then some marketing guy decided to stick their name into it; so now itās Topaz Video Ai. Same difference. Except the forum topics are still called things like Video Enhance AI Discussion. (Probably so links to older posts wonāt break, I reckon)
That has the ring of āsolutionā to it. Iāve had to do the same for QTGMC (ādevice_id=ā) in similar circumstances.
Even if you request a beta tester
no reply
https://community.topazlabs.com/g/Video-Beta-Testers
Anyone know?
Can someone explain to me the difference between Proteus and Artemis? I feel like I get better results with the older version of Topaz AI what happen?
Iāve not yet updated my NVIDIA 527.37 driver but have installed TVAI v3.1.0.2.m in parallel to v3.0.6.
Running same video parallel in both versions with identical preset (Proteus Relative to Auto) give me the impression that the new Alpha is faster than the older one: 1.4fps versus 1.1fps. It seems not to depend on new NVIDIA driver. However Iāll check the new driver when processing of the remaining video in v3.0.6 is finished in a couple of hours.
3.0.6, when running Stabilize on GoPro footage, I can preview (and it looks great) but when exporting to ProRes, 264, 265 (AMD or nVidia) the output video is a strange grid - see attached
GP010165_tvai.zip (3.0 MB)
Small video of the first few dozen frames in a ZIP
Settings:
Stabilise (AI)
Rolling Shutter Correction OK
Strength have tried 10-30, same result for all
INPUT Video:
GoPro Hero 4 Silver - 1080p48 Linear, directly off card
Machine:
32GB DDR5 6400
AMD Ryzen 7 Pro 6850U
nVidia 3080Ti 12GB
Windows 10 Enterprise
sadly none of the issues have been resolved. the preview playback never plays back as it should, you have to go to the file itself to pla it, and sadly even then its cocked up the speed of the video.
so just did a test preview with another codec away from prores 422 hq and found that its prores422 hq thats causing the severely stuttering slowness of the footage.
i would like to see 2 and 3 pass on settings such as the anti aliasing, so that it saves having to re import the first exported file and saves file space.
especially if the user has to upscale just to have the setting actually become active
No need, weāre all Beta testers now since V3.
lol
Reason
Because I deleted the early version
itās for getting
I have that exact same problem as well. I didnāt have it on 3.0.5. I am using a iMac Intel, Monterey 12.5.1
Radeon Pro 580X 8 GB. 32G Ram
After installing NVIDIA driver 527.56 no change in performance: 2.0fps with Proteus Rel. to Auto; 1920x1080 50fps to 3840x2160 50fps; H265 Main10 (NVIDIA), mp4; 35% GPU load; 25% CPU load.
Artemis is more of a one click solution, just pick the model that matches your source. It either looks good or it does not.
Proteus, I believe, is intended to be the jack of all trades. It started with manual setting sliders for fine tuning, then they added auto relative to slider, then full auto.
They should still have a tutorial video on how to best set the parameters in Proteus Manual. If Auto donāt not come out good, look for that video and use what you learn to get a better result.
All that said, I still think the only way you can truly get a full movie to look itās best, is to split it into scenes and find the best model or parameters for each scene. Yeah, too much work the casual hobby. Though I think thatās what Proteus Auto strives to solve. Maybe in the future it will become really good at doing that.
It wouldāve been nice if one could mark separate segments on the timeline and then assign different models/parameters to each segment.
I also notice that TVAI still canāt process variable frame rate sources. Can this be fixed?
Thank you for your reply. The quality isnāt the same as the past. I would take a 540p and upscale to 1080p. Then I would try to fine-tune it by zooming into the video and see if the AI model would smooth/sharpen the lines. Sometimes I think you would have to do a couple more passes in order to get the results you want. Mind you, these are 4-5 min long videos. Like you, Iām just a hobby and love to watch high quality videos.
Bad to know but good to know at the same time.
Youāre about as far away from me hardware and OS wise as is possible, so based on that Iām fairly confident itās a bug in the handling of the stabilise model in 3.0.6, and not something GPU or driver specific.
I have confirmed the latest alpha build also does not solve this and have reported it in the respective beta testing thread. Hoping it can be tuned up.
