Some bugs I have experienced seems to be fixed in 3.0.1, thanks 
I also like the UI enhancements I this version.
I still have this split view issue when previewing: the left image is stuck.
If you want to test, the âRĂ©siste (DVD).m2tsâ file is already in your Dropbox folder.
Thank you 
VEAI is no longer the name. Itâs now TVAI.
Yep, theyâve shown no interest in working on improving the models, which is all that will help this.
Think of all the time lost that those models could have been training WHILE the UI was being worked on.
nal_unit_type: 62(UNSPEC62), nuh_layer_id: 0, temporal_id: 0
Output frame with POC -510.
Decoded frame with POC -510.
cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
nal_unit_type: 1(TRAIL_R), nuh_layer_id: 0, temporal_id: 0
nal_unit_type: 62(UNSPEC62), nuh_layer_id: 0, temporal_id: 0
Output frame with POC -508.
Decoded frame with POC -504.
Decoded frame with POC -506.
Finished processing frame alq-13 0 0.008333 0.008333
w:4320 h:7680 fmt:bgr48le sar:0/1 â w:2430 h:4320 fmt:p010le sar:0/1 flags:0x200
Error encoding frame: -12902
popping: -542398533
Error initializing output stream 0:0 â Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
Uninit called for alq-13 0
Statistics: 0 bytes written, 0 seeks, 0 writeouts
Qavg: 243.692
2 frames left in the queue on closing
Statistics: 2231012 bytes read, 55 seeks
2022-10-26-11-08-19-Main.tzlog (300.2 KB)
Deinterlacing improved, but still a problem.
Applied the Interlaced/Progressive Dione Robust. It does a good, but slightly noisy job on a segment of old SD video, and a bit of haloing, too. The image is sharper, clearer and much improved when I switch to Interlaced/Progressive Dione Robust Dehalo, BUT there is too much decompression being done as whole image distorts inappropriately.
beneath_the_12-mile_reef-Interlaced_downloaded_144039039_dtds2.zip (4.4 MB)
Additional Problem: GUI Stability. If you attempt to change the GUI view from preview to input while the screen is busy, the GUI will crash. - This has happened twice in the last 10 minutes of useâŠ
Those are caused by âRecover Detailâ and âSharpenâ.
As before, Iâd advise to turn those to 0, switch the parameters to manual.
Since these artifacts are âonlyâ with some âdetailedâ faces, but not with other detailed objects, I think that is matter of better training of AI models handling faces. Recover Detail and Sharpen set to 0 will get iron smoth across every thing. Thatâs not the best compromise.
Ols male faces are ok.
Yes, this is complex. If you are exporting a file (you must declare it an export before starting and it will show up in the lists when you click on the âExportsâ button) it will be located in your âExport asâ file folder with the suffix _number_tvai as an mp4 file. If the file is a preview, it will be in a special internal folder and will show up if you click on the âpreviewsâ button. You can clear the previews by selecting âFile â close all sourcesâ (I currently do not have any previews. I think there are separate options if I have previews). You can control these from the File dropdown. You can see / set the default directories in the File â Preferences â Directories options.
Wow! Iâm not the only one. I saw the exact same old lady wrinkles on Janewayâs face in Voyager S01E01. I closed it up, shut it down and walked away shaking my head. And that was after a 24 hour upscale. My free upgrades end in two weeks, and Iâm wondering if its worth it or not.
You can install the geforce driver just normal.
The only thing you will miss is ISV certification.
And the driver maybe has more bugs. âmaybeâ
yes, the support of the cloud computer iâm renting told this earlier today, so iâm on a studio geforce driver. i didnât know that. it will be acceptable until a new driver will be released for the A4500.
I think the next pro driver will be out in mid November or with the release with ADA6000 in Dec.
ok, thanks, for now, i spent the day on it (but not on VEAI), driver works ok with the card, but didnât tried yet on VAi 3.0 or 2.6.4
One thing I will mention from personal experience, so not for everyone, is I have always found Proteus to be more of an âin houseâ avisynth style tool than a proper upscaler. I have sunk countless hours into trying to replicate the details in faces I can get using other models and versions in Proteus, and so far, no success, no matter which version of Proteus I play with.
It seems to be more tooled for cleaning up footage than upscaling, but almost universally for me, produces artifacts that seem more in line with scripts being used badly moreso than the AI upscaler not working. Those âwrinkleâ lines I have seen appear as well, but its always on images being over sharpened and usually the result of the original having leftover interlacing issues that are not removed before being pushed by scripts into exacerbating the appearance.
Personally, I have found Proteus to be the worst of the models when upscaling, which for me specifically bodes not so well when I see it become the default selection tool in the application. But this just may be that they need to go back and do a lot more work on this model. The only downside is they had models that worked a lot better, but the attraction and more simplicity of being able to customise some of the settings in Proteus, over changing the input source seems very strong.
There is a parameter in Proteus that I donât understand.
2.6.4 Manual
Only âAntialias / Deblurâ can be set to a value from -100 to +100, while other items are set to a value from 0 to 100.
3.0.0 Manual
Same as 2.6.4. (For CUI, set the number to 1/100 ex,100 = 1,10=0.1)
3.0.0 Relative to Auto
All settings can be set from -100 to +100.
The help text for the veai_up filter is as follows
âestimate Number of frames for auto parameter estimation, 0 to disable auto parameter estimation (from 0 to 1e+06) (default 0)â
Q1
For example, in 2.6.4
Revert Compression 20
Recover Details 24
Shapen 0
Reduce Noise 16
Dehalo 0
Anti-Alias/Deblur 4
If the CUI settings in 3.0.0 were
compression=0.20:details=0.24:blur=0:noise=0.16:halo=0:preblur=0.04 (estimate=0 or not set)
Is this correct?
Q2
I am not sure how âRelative to Autoâ works.
For example
Revert Compression -30
Recover Details -10
Shapen 0
Reduce Noise 10
Dehalo 0
Anti-Alias/Deblur -5
The value automatically determined by TVAI is internally
Revert Compression 20
Recover Details 24
Shapen 0
Reduce Noise 16
Dehalo 0
Anti-Alias/Deblur 4
Relative to that
Revert Compression 0
Recover Details 14
Shapen 0
Reduce Noise 26
Dehalo 0
Anti-Alias/Deblur -1
Is this a function to?
Q3
In GUI, estimate is locked to 20. What is the effect of changing this value in CUI?
Hi, thanks for your continuous effort on developing this tool.
I have found a couple of issues that are disrupting our current workflow:
When rendering to a tiff sequence, the folder name now includes a random number. What is the reason for that? Wouldnât it be better to keep the old behaviour of adding a â(1)â suffix when the same file is processed again?
Also, the tiff files are now zip compressed, making the reading a lot more difficult in editing software. Could you please revert back to uncompressed tif files, or at least give us that as an option?
Thanks!
Hi there
I posted this questtion yesterday but just deleted it as I discovered something else that was going on in the background, so here is my re-edtited question from yesterday:-
I just downloaded this update earlier and selected Export to upscale a video to HD which it began to process, I then selected another video from the same source to do the same again, but nothing happened. I tried a few other videos but nothing was happening.
I then checked Preferences but it was on the setting for converting 3 videos at once, so I then closed the software and rebooted my PC, And then when I tried the software again it would not convert anything at all!
So then I uninstalled the software and reinstalled it and did the very same with my latest Nvidia drivers, rebooted again, but itâs still not letting me do anything at all!
Can anybody help me???
As an addition to my question above I have discovered that the software is actually doing itâs work in the background via FFMPG. However, I am unable to test my settings beforehand to make sure I have each conversion set the way I want to with Preview, since the Preview window wonât open at all?
