first issue: did 2 previews of 10 seconds on different points, the software considered both previews as a single file and the 2 one would not play at all. will play the first preview as if it was the 2 preview.
actually it will preview over and over the same spot of the video. i am exporting a short video to see results. selected proteus and all values from the stimate pushed to 100. (just for testing purposes)
will edit/reply with results but its more broken than 4.0.5.1.b
edit:
now apparently the software is 0 seconds remaining processing over 4 figures frames per second?
i did not know i have a quantum serverā¦ way way broken like unusable build
edit 2:
finished the export at random, and ignored my proteus settings. exported the video untouched. totally broken unusable build. stopping testing since its a waste of time.
Unfortunately, when I close a video file in Video AI, I can neither move nor delete it in Windows Explorer. I have to open another Video or close Video AI completely to do this. Unfortunately, this has always been the case. Is there a technical solution to this problem?
Well, but this for sure is not specific to this Beta but a general problem with the Iris model.
Iād suggest to use Iris LQ and high Recover Original Details if 85 to diminish that effect.
You missed my point: It is terrifyizng that TVAI IRIS invents any artificial details where there are none in the original just because AI interprets a shape detected in context of a person to be likely a human face.
Up to now we complained only about artifacts and deteriorated details enhanced by IRIS, but not about complete replacement by any AI generated face.
Off course Artemis or Proteus are the right answer in this case because they donāt invent details which are not there.
Part:
Yes, they do - just not with faces which is why you donāt notice it as much. Besides it kinda is the task of the AI to insert some missing details otherwise you could just do a bicubic upscale.
And, of course the AI (I dont really like the term cause IMO thereās no real intelligence in there, just learned/programmed scenarios) sometimes gets things wrong.
When you choose either Nyx V1 or Nyx V2 and attempt to set a Recover Detail value, the value resets to 0 or the previous value if you enter a value by typing the number directly. Additionally, if you attempt to save a preset for Nyx with a value set for Recover Detail it will reset to 0 or at least be presented as 0 when you load the preset.
@luke.roche Topaz Video AI Is Down Mixing or Striping to 2 Channels. A Dolby or Multi channel Video Source would be useless in this program. Channels would be lost .Program needs At least Spacial Audio for 2 channels .My Source file was 5.1 Multichannel. From Video AI now 2.AC3 still mixed down to 2 channels. Cats1MediaInfo.txt (4.1 KB)
FFMPEG command here
The absence of Linux builds for newer versions and the auto-closing of threads means Iām struggling to know where to report apparent random memory allocation failures for the 3.5 TVAI. It seems that something decides to try and allocate TBs of memory on my 8 GB RTX 3070 Ti cards, and retrying a few times will eventually allow something to go through and run.
?