Hello! I am using VEAI v2.1.1 and running into a frame duplication/stuttering issue with all models and output methods. I tried changing the VRAM setting but it had no effect. The input video is 720x480, progressive, 29.97, ProRes and I am seeking to output at 100%. VEAI outputs a 654x480 video, which I thought could be related, but the stuttering happens with a 640x480 input video as well. Do you have any suggestions for how to troubleshoot this further?
Does the frame stuttering see the same frame display on screen go back and forth by what looks like a single frame while the frame counter continues to go up, and then when you cancel topaz closes down completely?
Yes, the frame duplication occurs in the VEAI interface before I send the video. Checking the frames in Resolve, QuickTime player, etc. confirms that the source file does not contain duplicate frames. Thanks for the tip. I had not actually checked to see if the frame duplication happened in the Topaz interface, I was only looking at the output files. I have not experienced the app crashing though. Do you have any suggestions for how to investigate this further?
As I wrote on an earlier post, If you have Bitdefender Antivirus running, go into the ‘Ransomware remediation’ settings by clicking manage, and add an excemption for the Topaz video AI exe. I was experiencing repeating frames that would not progress, resulting in the loss of all progress.
Thanks! I’ll look into seeing if some other software is interfering with VEAI.
Is it better to overwrite an old installation of VE AI with a new one? I’ve always uninstalled before installing the new one.
Sorry if this seems like a dumb question. I’m just curious.
Edit: I didn’t know the setup provided a “Clean installation” box that you can tick. Incredibly handy.
Good news: you can use older versions of AHQ! The trick is to go to your models folder (on Windows usually: “C:\Users%USERNAME%\AppData\Roaming\Topaz Labs LLC\Video Enhance AI\models”) and make a copy of the latest AHQ .json file that comes with your version of VEAI – right now being artemis-hq-11.json. Rename the copy you just made to the version you want to download, just to keep it organized, which should be artemis-hq-5~7.json depending on which you prefer (I think v1.0.1 corresponds to 6). Edit the contents of the file and change the 3rd line, where it says “version”: “11” to “version”: “6” (or 5 or 7; you can have all of them in separate files). Now close VEAI if it’s open and open it again – the new model will be in the list 
1 Like
To upscale / enhance my HDV material i am using pre-de-interlaced files with doubled framerates and created a progressive version of Dione Digital Video V2 to get the best results. The build-in Dione Interlaced Video V2 produces massive flicker in most HDV files, especially if you have horizontal movements.
if anyone is interested in trying this out you can download the files and instructions here:
http://www.tochti.com/public/dione-dp-2.zip
There is also a Windows BATCH file for ffmpeg included which i use to de-interlace my HDV files before using them in VEAI. Important: This batch file can only use NVIDIA NVENC !!
1 Like
strange. Do you render in mpeg-4 or in frame sequence?
I render in frame sequence and then compile everything with Media encoder in H265. I updated to Video Enhance 2.1.0 on 3/25 (the day it was released) and I started getting blue screens since last Sunday or Monday. And I had 2 more yesterday.
Most of the time, it is during the final compilation. But it is always after using video enhance at least 1 hour before. However, when I look at the temperature of the GPU, during the final compilation, I am below 60°. With Video Enhance I am at 80°. To maximize the space on my internal disk, the software is installed on an external disk. The export in image sequence is also on this same disk. This disk is 2TB. And with my softwares installed + the image sequence folder, I have 816 GB free.
I also created a registry file to move the templates to this same disk. But maybe the registry key I created on 2.0.0 is not valid anymore for 2.1.0 and 2.1.1.?
I have uninstalled completely and will reinstall by manually recreating the registry key. Maybe it will be ok. In which case I’ll switch back to 2.0.0 and hope that the 2.1.0 templates will work on it.
I always render in ProRes which is the most convenient format for my purposes (we are remastering source footage from a feature film we shot in 2004 in SD).
Could it be that your external disk is causing this? Maybe the USB3 controller overheating or something like that?
i am going to disable resizable bar in my mobo bios.
but i hope topaz fixes this issue.
also my 3090 has 24gb of vram and this program neve utilizes anything close to that limit.
This morning I uninstalled the software, then I made my compilation which had crashed twice yesterday with bluescreen. And this time I could do it. Then I restarted my PC and reinstalled video Enhance. We’ll see later if I still have the problem.
On the other hand, I notice that with each update, there is a trace of the old version in (control panels/all programs). I saw the 2.0.0 and the 2.1.0. Every update causes this. The installer is not as good as the other Topaz software.
Another point, since 2.1.0, I had noticed that the button “load last auto save” did not work anymore. I had to re-import the sequence each time and adjust where I stopped
Great updates! Thank you!
Some requests:
- We need the ability to PAUSE processing.
- We need the ability to ADD / REMOVE videos from the queue during processing.
- We need a “test mode” where we can choose 3 or 4 different algorithms at the same time (5-way vertical split of the screen during preview with different algorithm used on each column)
- Please change the sec/frame system to frames/second (FPS). Nobody uses the sec/frame method. Seriously.
And a happy discovery! (may be helpful to other users):
On my dual Xeon system (2x E-2670) with a single RTX 2060 SUPER and GPU acceleration, I get 0.15 - 0.19 sec./frame using Artemis LQ.
When I run 2 instances of VEAI, I get 0.17 - 0.21 sec./frame on EACH! This means I can process 2 videos at the same time with a little loss on speed!
To make that clear, for a 50min video I usually need 2.5 hours.
Processing 2 videos at the time (2 programs running) I need 3 hours.
Result: With just an added half hour on processing time, I get 2 videos instead of one!
On a regular day, I can get 9-10 videos done with one program running.
Now, with 2 instances running on the same computer at the same time, I get 16 videos per day
Keep up the good work!
P.S. About the pricing: We all understand the need for new customers. And we are ok with all those extra discounts etc. But you should at least offer something to early customers. Like a few months free expansion to updates or something. We paid full price back then and new customers pay only a third of it.
It’s not fair for the early supporters, just saying…
4 Likes
i guess with two instances running at the same time it should also be possible for the developers to implement this feature in one application during queue processing. Could be an option in the setting to define up to 4 parallel encoding processes running at the same time and a small display of the numbers to switch the preview window.
I think depending on the resolution of each clip it would also be possible to roughly calculate the VRAM usage for each process.
1 Like
After the 1.2.0 update, I noticed that the GPU utilization dropped to 34% on HWinfo64 monitor, without any speed loss on processing.
At first I thought it was a bug or something, but just for the heck of it, I tried running 2 instances of the program, thinking that 64% if the GPU seems free.
The idea behind it was like “2 threads” of a CPU working on different tasks. And it worked!
Let’s hope the developers find a good use for it.
i take this for a warning > NOT to update my RTX3090 bios soon
it is anyways a good idea this days, as we are all aware new rtx 30x0 gpus are hard to find and horrible expensive, to treat our cards like an fresh egg… replacing, even on waranty, is almost impossible right now. IMHO
expecially overclockers should take care
for me (rtx3090) its just vise versa…
before 2.1.0 i had bsod regulary, during multiple veai instances rendering,
since 2.1.0 they are gone …
but at the same time i switched from adobe premiere, coz of their bad rtx3090 support=extremly slow exports, to davince resolve 17… since then bsods disappeared
system i9-10900, win10, 32gb 4266 ram
i just had some strange issues after installing 2.1.x while importing footage and seing just a small square in the preview window, or crashing right after hitting the import button
first i thought my codecs are broken, but a fresh veai install … helped
my psu is an asus rog thor 1200W, this is extremely powerful and oversized, i know,
but for video rendering on highend cpus/gpus, you cannot use a cheap “china psu-bomb” for 50$
always keep an eye on your cpu/gpu temperature…
once they hit the 100° C → they can shutdown/bsod every second…
confirm, in taskmanager i only see 7-15% gpu load since VEAI v2.1.x
but performance is like before
so i now use an asus gpu monitoring tool, so i see even when taskmanager shows 7% only…
real gpu load is around 50% or more
This has been a problem for a while and is scheduled to be fixed in the May release.
@taylor.bishop As far as nvidia’s 3000 series gpus are concerned, VEAI simply refuses to work with the latest 465.89 drivers (which introduced resizable-bar support, among other things.) for some reason and gives the error “something something this model failed to load, reduce vram amount and try again etc.”
The only way to make it work, albeit obvious incompatibility, would be to tick “reduce machine load” and only do 100% deblock/denoise. But even then the process crashes well before it’s done enhancing, throwing the aforementioned error.
Reverting back to the previous 461.92 drivers seems to solve this issue thankfully, but having to switch back and forth between drivers for gaming/video enhancing purposes can be tiring at times, so it’s by no means a proper solution for the long run.
Me and some 3000 series gpu owners (@eternalkp and @mxrevolution so far) are all experiencing the same annoyance unfortunately so we kindly hope you guys can come up with a fix if you please, before even more people start facing this problem as well.
1 Like
