Are there any improvements on Artemis HQ? 
I didnāt see it mentioned on the notes, and if not, this should be the second release without any improvement on HQ. So technically using HQ10 vs 11-12 should be the same?
Iām asking mostly for disc space I donāt see much reason to download all the files (from HQ) from the new version if itās the same. Thank you.
it seems that dione tv v3 cant be run on older cards, even lowering vram
My initial experience in going from v1.9.0 to v2.1.0:
On the minus side, VEAI now skips most of the frames of a .VOB input file, giving a 3-minute output file from a 1-hour interlaced input file. This is using the Dione interlaced models.
On the plus side, I deinterlaced the .VOB file with Handbrake and saved as an mp4 file. Using this as the input file for the Artemis LQ v12 model gives great results while correctly processing the entire file.
The file is a terrible quality video from a primitive digital video cam 27 years ago (my wedding). The Artemis results are much better than I had hoped.
Weāll be fixing this (and a lot of other video I/O issues) in the May release. Thank you for the report!
Was this not an issue in the v1.9.0 release as well?
Is there anything in progress regarding some drop of performance after processing 4-5 or more source files in a batch? It has been there in 2.0.0 and it is currently going on in 2.1.0 I upgraded to yesterday evening.
The performance after 4-5 files in a row is dropping progressively. For the first file itās ~0,31seconds per Frame, with the fourth or fifth file it starts dropping to about 0,4-0,5 seconds per frame and after the 10th file I canceled the process in 2.0.0 where it reached more than 1,5 seconds per frame. Thatās while using AMQ/ALQ v11/v12 on episodes of an older TV series upscaling to FHD on a fresh 20H2-system with the latest Geforce-driver installed. No problems regarding temperatures or power supply. In contrast: In GPU-Z you could see, that the GPU feels more boring the more files were processed.
When I want to start over I could cancel the process, but when I try to close VEAI it eats up to 32GB memory (whatās in the machine) and I have to kill it via task manager. I simulated the problem on an totally different 2004-system under v2.0.0, too.
HQ is at v11. MQ and LQ are at v12.
As always, it all depends on the source:
Here a Super 8 clip from 1964, filmed by an amateur camera and a correct lens but not an Angenieux ā¦
The film has a very coarse grain (Kodachrome 25 from memory) and it becomes very difficult to keep all the details and cleaning in preparation for an edit.
The Artemis Strong Dehalo V1 model managed to do this straight away even at 200% retaining all the details. Even better than AMQ / ALQ : depend the source
Search for how to make symbolic links in Windows. Then you can move the models folder to another disk and link it back to where VEAI expect it to be. Once you learn how, itās very easy.
Soon after the FPS conversion model is added and video I/O issues have been addressed (May release). Our backend doesnāt currently support a high bit depth per-frame, so we need to have that sorted out before we start on SDR ā HDR.
Strangeā¦ No Tiff 16bit uncompressed? Itās still an option in 2.1.0.1bā¦
@taylor.bishop No Tiff 16bit uncompressed as an option for output?
Is this option not showing up for you? Can you send me a screenshot of the Tiff 16bit uncompressed option missing from your interface?
None of the output format options should have changed in this update.
Great work, guys - fantastic update! The GUI seems a whole lot snappier now, and doesnāt take a fortnight to start up. I really like the new timeline selection controls.
Also, the new Dehalo model is working wonders on some problematic videos that Iāve been trying to fix.
This is amazing.
hi sir.I have just updated my video enhance to 2.1.0.but the language is strange on the interface.no letters or words only symbols ā¦please any help to solve this issue?my operating system is M1mac.the previous version was working right
OpenVINO will not work on the 2.1.0 release version of VEAI.
This error occurs:
In checking the logs, I found this line:
2021-03-27T00:49:01.409 [Debug ] < line #: 0> Thread ID: 0x1c022caf650 - Running CPU fallback because of this error [ Failed to create plugin MKLDNNPlugin.dll for device CPU
I checked the program directory and sure enough, MKLDNNPlugin.dll is missing.
This file is present in Video Enhance AI 2.1.0rc2, where OpenVINO works fine.
If I copy that file into the program directory of the final release version of VEAI 2.1.0, OpenVINO then works.
Here is the file in case anyone needs it:
Ā
Another thing to note:
Starting with Video Enhance AI 2.1.0rc2, it no longer starts on a GPU-less machine without further user intervention. It throws this error:
So VEAI versions 2.1.0rc2 and above now need Mesa. This is the first time ever that VEAI needs intervention from the user regarding Mesa in order to run immediately after installing on a non-GPU machine.
There is a very old version of Mesa that the VEAI installer copies into the VEAI application directory as opengl32sw.dll. If itās renamed to opengl32.dll, then VEAI will run.
The latest version of Mesa that I recommend is MinGW 20.3.4. Versions after that (so far) have unresolved issues. For Mesa version 20.3.4 (as well as other newer Mesa versions), both x64\opengl32.dll and x64\libglapi.dl need copied into the application directory.
I havenāt checked for sure what environment variables it needsā¦ by default, Mesa runs OpenGL at 3.3, but I have mine set to 4.2. I havenāt tried deleting my environment variables and checking to see if VEAI runs OK without them.
(Simple Classic Theme for Win10 in case anyoneās wondering. 
)
hi everyone.I have just updated my video enhance to 2.1.0.but the language is strange on the interface.no letters or words only symbols ā¦please any help to solve this issue?my operating system is M1mac.the previous version was working right
Tiff 16bit uncompressed not showing up but itās still here 
however not easy to see in a 1080p monitor because itās the last option. I have to change my task bar settings to hide the task bar in desktop mode.
Hello, I noticed that the Artemis models medium and low version 12 are slower than the version 11. Before I was at 0.23-0.25 seconds per frame. With v12, I am at 0.31 seconds per frame. This is not normal.
That sounds like the font issue that a few others have reported.
thank you lee.do you have any idea how to resolve that? please
