Hum, I see
Why not wait a few days and see?
You do not need a Facebook account to participate in the betas. You need to join the beta testers group HERE in this forum. The links will be placed here, and the discussion takes place here. The other person that replied to you is a bit behind on things considering that they havenāt got the āBeta Testerā flag on their name.
This is the link to join the beta testers group HERE, which is the only current place for beta testing.
https://community.topazlabs.com/g/Video-Beta-Testers
@plissandro - You should do the same.
Yes , there would be an easier way to directly see the MPG-2 stream colorspace info using MediaInfo on one of the DVD VOBs but the encryption on commercial DVDs prevents MediaInfo to access the file.
Are you sure this info isnāt specified on the MKV by MediaInfo ? Your MediaInfo screenshot seems cropped.
they are posting the links to download the beta versions on the facebook aswell, so joining ābetaā is needless unless you only care for the ābeta terterā title here.
Also Iām not interested in beta testing VEAI
Itās being transitioned here. All discussion has already moved here.
Please make Gaia HQ good again and improve it much further, please.
Please further improve Artemis LQ so it does not blur details, but nicely eliminates blocks in videos.
You really needs to train your models on people. Most scenes show people walking or doing something. Please make models of people in clothes and nude, countless videos.
Gaia HQ adds that black noise on human body what looks bad.
1 Like
Ahhhh, if Artemis MQ upscaled faces as well as Gaia HQ does, we would have an āalmostā perfect model.
I put all the info visible on the media info.
I just recompiled the episode in Handbrake by forcing the 25 frames/second at the same rate as the original MKV file and I have no loss. And now I only have 25 frames per second in Video Enhance.
I just bought RTX 3090 as replacement to RTX 2070s, but the result was extremely disappointing:
RTX 2070s (1080P to 4K)
Artemis LQ V.10 : 0.43/sec
Gaia HQ-5: 0.62/sec
RTX 3090 (1080P to 4K)
Artemis LQ V.10 : 0.43/sec
Gaia HQ-5: 0.61/sec
itās werid that the speed is almost the same and no different at all~~~~~~
why~~~~~
VEAI models do not take full advantage of the 3090. Run two instances of VEAI on a 3090 for good performance.
3 Likes
I think this is also due to the Ampere architecture, which is made for raytracing, Turing was compute and raytracing.
From my tests I found that the 3090 behaves as if it had 20 teraflops.
I would be interested to know if there is a difference in the new Pro RTX cards, as there was in the previous generation.
Update: Read this before my Bugreport: The Bugs make sense somehow, my images donāt have a Timecode. Maybe thats the cause of the Bugs.
1.8.1 Bugreport: When i try to drop in 382 images (8 bit Tiff files ) the Software does crash.
27 images was ok. ā> 41 also, 55 not.
Appens also with png files.
Why do i drop in so less images?
Eh, i try something.
Update:
Another bad bad bug.
When i drop in, the pngs (41) mentions above, and set up Theia FI-V4 Sharpen 50, Deblock 50 and Noise 50, 100% Denoise Deblock (checked fill frame), Output PNG 8 bit.
and then after Setup i press Start it does export this (Noise-2021_2.00x_7776x5184_ghq-5) to a different folder.
Did try this 4 times.
Maybe its happen because i did try to process photographs instead of a video, but with artemis it does work.
2021-01-03-14-01-7-Err.txt (1.7 KB) 2021-01-03-14-01-7-Main.txt (38.0 KB) 2021-01-03-14-01-7-QML.txt (4.6 KB)
Another Bug:
When i drop in a single png file from an folder containing some files it does show the single image in the UI, but when i press Preview (to show whats happen on a single file) it does load the next 30 images also.
When I enable āuse all gpuā processing it gets stuck in the āinitializing AI engineā loading screen forever.
My specs are:
i5 4690K
GTX 970
16 GB RAM
Z97 motherboard
128 gb ssd and 1 TB hard drive (on which I export the video)
Running the latest non-developer version of Win 10 with all the lastest drivers.
Please resolve this,
it would also be amazing if you implemented in some way intelās integrated graphics QuickSync technology like in Apple computers, that would extremely boost the rendering speed and efficiency!
(In conjunction with NVIDIA NVENC)
Soā¦Iām afraid to try these new versions since 1.6.1 works well for me.
BUT
It looks like I only have openVINO and TensorFlow models in my models directory. I followed some instructions to download models, but I canāt seem to get it to work for meā¦I always get an error page.
Is there a way to get all of the older models that 1.6.1 used? Or am I screwed and forced to upgrade the entire program? See, Iām still on Windows 7 on my VEAI machine which is NOT connected to the internetā¦so if there are models it needs to download, it canāt. The machine is running a 1650Super btw.
Thanks for the great product, and thanks for any help!
NVENC would destroy most of VEAI quality outputs and would be completely useless with the computing rendering images per seconds you get. On the other side x265 encoding with slow presets would achieve great results and saving time. NVENC is meant for speed in trade of global quality. Why on earth would we need NVENC encoding in a program that can render 0.1 to 5 frames per seconds at the best ?
People Like API Words.
CUDA, QUICKSINC, NVENC, OPENCL 1.0, 1.2, 2.0, 2.2, VULKAN, DIRECTX, DIRECTML, PHYSICS.
Pick one.
I like them too, you always have such a great idea of how everything will be better if the software you use makes use of XYZ that sounds great and someone did make a lot of advertisement for it that also looks great.
The thousands of lines of code behind it you unfortunately not see, and that someone has to build it and it fits together to the idea they did have , drag and drop does not work here.
1 Like
