I have very poor jpg quality on VEAI 1.7.x.
It makes 1/10 file size jpg of VEAI 1.6.x and have very poor image quality and got much bending.
on previously versions of VEAIs, I have almost no visible differences of quality on JPG and PNG, but in 1.7.x it is very not.
jpg output of VEAI 1.6.0 (part of 7680x4320 image. file size 6,299KB)
though PNG and TIFFs are lossless, but it has very huge file size. 2-hour movie 2k to 8k, about 16TB for a 8bit TIFFs, 34TB for 16bit TIFFs, 6TB for PNGs
and PNG output is very slow. it is almost double time slower than jpg or TIFF output.(Iâm using RTX3090, 0.85sec/frame of JPG or TIFF output. 1.5~1.6sec/frame for PNG output. 1920x1080 to 7680x4320)
I think it is cause of VEAI using CPUs for PNG endoce, and it is very slow on 7680x4320 PNG compressâŠ
Can you guys make VEAI get back to 1.6.xâs JPG quality? or maybe you can make Quality Option for JPG output (like CRF option on MP4 output)
I was wrong. The Windows Firewall still blocks some of the features. I have accepted everything. Preview/process doesnât work unless I turn off the Windows Firewall. Half of the preview is blurred (must be related to downloading models).
Few options: 1. Reset Windows Firewall (Windows says it might stop programs from working) 2. Install Tinywall 3. Wait for an update?
I had a problem with immediate reboot when VEAI initialized the video card. Turned out it was power supply, but not wattages. It was the amps - way too low. High wattage PS donât necessarily have a step up in amps for video bus. In my case, EVGA 850 watt have more than enough amps.
If anyone else is building a second VEAI PC, one license is good for two PCs!
General FAQ - Topaz Help
âHow many different computers can I install my software on?
The software license for each of our apps is good for up to two different computers.â
I just bought another used GTX1080Ti from the same seller on eBay with Buy It Now. Price was the same at $550 + free shipping (returns/replacement offered too). I was going to snipe an auction or 3, but bids for several cards were over $500 three hours before auctions end, not including shipping and they didnât accept returns. I was planning on testing 1.7.1 but several of you saved me the time and trouble by posting the block artifacts, thanks!
Version 1.6.1 is going on it for now. I hope Topaz can fix the issues as it would be nice to use the new much faster and more efficient GPUs. They should recall 1.7.1 until the issues are fixed and go back to 1.6.1. If it was my company, thatâs what I would do (and Iâd have a serious discussion with the developers).
Preview/Process would require internet if 1) âDo not download faster modelsâ box is not ticked, AND 2) a particular model file has not been downloaded yet.
Not true. Before replying I recreated the default conditions. I removed all LQ V8 files except the 288x288 files that come with the installer, blocked internet access to VEAI (actually my normal running condition) and then loaded up a big file and ran it (Do not download NOT ticked). Running through the log:
2020-12-02T19:32:33.894 [Info ] < line #: 0> Thread ID: 0x1c5f1fdd730 Locating local model: âC:/Users/xxxx/AppData/Roaming/Topaz Labs LLC/Video Enhance AI/models/alq-v8-fnet-fp16-480x384-2x-ox.tzâ status: false
2020-12-02T19:32:33.894 [Info ] < line #: 0> Thread ID: 0x1c5f1fdd730 Fetching remote file: alq-v8-fnet-fp16-480x384-2x-ox.tz
2020-12-02T19:32:33.920 [Info ] < line #: 0> Thread ID: 0x1c5f1fdd730 RESULT for http://veai-models.topazlabs.com /alq-v8-fnet-fp16-480x384-2x-ox.tz is 0
Repeated for a bunch of other sizes, then:
2020-12-02T19:32:33.923 [Debug ] < line #: 0> Thread ID: 0x1c5f1fdd730 Loading default model file: C:/Users/xxxx/AppData/Roaming/Topaz Labs LLC/Video Enhance AI/models/alq-v8-fnet-fp16-288x288-2x-ox.tz
And finally:
2020-12-02T19:32:34.695 [Info ] < line #: 0> Thread ID: 0x1c5f1fdd730 Loading time for model file: C:/Users/xxxx/AppData/Roaming/Topaz Labs LLC/Video Enhance AI/models/alq-v8-gnet-fp16-288x288-2x-ox.tz is 435971
Under the same circumstances, I get an error message and processing did not proceed until I ticked the Do not download. That is, unless the source video is of a resolution that would use the standard 288x288 tile anyway. Was your source video 480p or smaller? Also, I rarely use Artemis, usually Gaia HQ.
Iâm always willing to test, so here is ghq with the same circumstances except this time source was 1920x1080. First is the line where it checks and finds that downloading is allowed:
2020-12-02T20:47:27.925 [Info ] < line #: 0> Thread ID: 0x26a2e3e2550 RESULT for http://veai-models.topazlabs.com /ghq-v5-gaia-fp16-576x672-2x-ox.tz is 0
Finally proceeds to default 288 file:
2020-12-02T20:47:28.807 [Info ] < line #: 0> Thread ID: 0x26a2e3e2550 Loading time for model file: C:/Users/xxx/AppData/Roaming/Topaz Labs LLC/Video Enhance AI/models/ghq-v5-gaia-fp16-288x288-2x-ox.tz is 878574
I tested my system with lower settings at yesterday, and can encode with any mode and any parameter for hours without any crash. My power supply have no any problem, but I think VEAI overload my video card, and if I decrease GPU max frequency and power limit, I can run without any crash.
Just out of curiosity: Howâs the performance with a NVIDIA SLI setup? I have two GTX 1080 on my shelf and an older i3 CPU with a dual PCIe Mainboard laying aroundâŠ
Itâs strange that mine fails. I did not examine the logs, but perhaps I will at some point. For now, I have no issues, as I made sure that I have all relevant models in the folder. The only other oddity I found during testing was that, while Artemis is usually faster than Gaia at 1x and 2x upscale, at 4x upscale it is slower, no matter the source. In general, I find 1.7.1 to be more stable than previous versions; it almost never crashes, even with mp4 output.
