Actually I wanted to serve 2 gods simultaneously Prometheus and Proteus which looks like Protheus maybe I annoy both Greek Gods
Today I was again able to select interlaced and Iris LQ in this version so probably the Iris Model got updated Yesterday it was only possible for Proteus.l
The result is phenomenal interlaced Auto and select on manual the estimate of the worst part with faces then select back auto to relative Faces Colour skin are Incredible.
The noise reduction and objects are fine as well smoothing is better
This is a definite gamechanger some of the Iris models must have changed since yesterday as interlace and interlaced progressive are possible
Again interlaced 50 frames output 2 X results in an incredible Export
On MacOS on a M2 Mac Mini, using Iris LQ and an upscale to 4k the app crashes after a minute, but ffmpeg keeps running in the background.
yes sir
The Iris LQ V2 does look good on close-up face shots but on fuzzy VHS quality distant shots you do get monster faces occurring here and there. I would like to see an Iris model that is just for poor-quality VHS sources that skip enhancement of faces for distance shots if it can’t get a good lock on them. A slider to control enhancement for eyes would be very helpful as well. The eyes often look over-enhanced in Iris mode.
Iris LQ V2 clearly better than V1. Faces in e.g. Video 8 videos can often be further improved if it is run twice in succession in normal PAL SD resolution without upscaling. The end result is then more precise than Iris MQ. I would be very happy about a new Iris MQ V2 version! Overall, the rendering of plants and cobblestones could be further improved in all Iris versions.
I can choose to download Aion in the model manager…
There are 2 files being downloaded.
But i can´t choose to use it for interpolation on windows.
Same here, yeah, I’ll wager it will be in the next beta/alpha hopefully.
Hi, finally I managed to test aion on 4K videos, by cropping at 3520x1980 adding stabilization, (crop mode from 20 to 50) which eases the work of interpolation task. It seems that in my computer it is the maximum resolution (or another one very similar) that the program allows me.
I have used several videos where Apollo V8 totally fails with tilings and other anomalies. Aion’s improvement is impressive, not a mere increase, it is an all or nothing qualitative improvement. In videos where Apollo and the rest of the models failed, I get now usable videos. It’s not perfect, there are still some subtle anomalies, but the model is on another level.
The problem is that it consumes a lot of memory resources, which increases above 10GB and sometimes causes the swap space to be used, which causes it to go even slower. In my case, Mac Book Air M2 24 GB, at just 0.1 to 0.2 frames per second, stabilizing and interpolating, leaving improvements and scaling for a later process. In all ways, the improvement is such that it is worth the wait for me (I usually process clips no longer than 60 seconds).
I don’t know why it just fails (red cross) when I exceed the mentioned resolution. But it should fail in a controlled manner, warning the user of the resolution limits. It is not due to memory saturation because the failure occurs before ffmpeg begins to consume memory massively.
Even with these limitations on my platform, and probably on others, and correcting the red cross problem with a prior warning, I think Aion is an extremely useful model.
FFMPEG: Like others, I sometimes copy the Topaz Export cmd before editing output properties and past in new cmd window…
I also often use ffmpeg to correct video problem before loading into Topaz.
Is it possible to load video into Topaz and add a pre-topaz ffmpeg-process to cmd. So the first process task is running custom ffmpeg script, followed by the Topaz Export cmd and the video is only saved once!
If not, how about an intelligent AI light correction feature.
Preferably with limits sliders so we can fix isolated frame areas.
Infognition Intelligent Brightness and VirtualDub MSU Smart Brightness, Contrast Filter have such limiters but can also introduce a lot of noise or even staircase/edge stepping.
Personally think it should correct/adjust light levels, Not contrast!
Like making objects and details more visible in dark footage, or reducing glare from sun or bulb.
You’ve All Done Very Well
“”“”“”“”
tony.topazlabsPRODUCT SPECIALIST
1
Update on release timing:
In 2024, Video AI will be switching to a biweekly release schedule.
Our goal with this change is to allow for more bug testing both internally and with our public beta testers. We will be alternating between a beta and main release each week.
4.1.0 will be released on Tuesday, January 16th, followed by 4.1.1.b on Tuesday, January 23rd.
“”“”"
In the last beta version topic, it was written that the new testing version would be released on January 23rd. Today is January 24th and I don’t see the new beta version in the beta section of the Topaz Video AI program. Is this the case for you folks as well?
Hello,
4.1.1.b will be released tomorrow, Thursday Jan 25.
We’re making a slight adjustment to the release day for betas to allow for more changes to be bundled together before posting for testing. Thanks!