I agree. I admit, I didn’t go near VEAI for a while with a lot of videos because the models were getting molded into shape but we’re reaching a good point where clarity is coming out out of de-noising, rather than de-noising sacrificing clarity. The deinterlacing models will improve, and I suspect in a few months they’ll be spot on. For now? They’re interesting and work in some cases. I need them to work in all cases and way beyond QTGMC much like VEAI’s Artemis LQ model has became my favourite de-noiser.
1 Like
They’re interesting and work in some cases. I need them to work in all cases.
I wish for that as well but I already scratched that a long time ago. 
Since like I said the damn Placebo-Option will not work just like it’s advertised I think it’s pretty hopeless. But if there are at least options now to go for the lesser evil that’s good enough I guess. Just like I said there is also some possibilities to just put in more work in “post-production” my fear with that is just that the next version of topaz VEAI or maybe even another company might make the whole work obsolete or even worse requires another approach
As writing that, I wondered have you tried downscaling your projects? A few months back I made some tests with it and had the impression that it looks somewhat better even with other models like gaia. The downscaling at least makes the artifacts less noticeable if even at all. Unfortunately it has not the effect I hoped for making a second run with VEAI. It’s almost as if the AI (even when a different model is used) can sense that “this is already processed”, since there is almost no effect, even when it is downscaled below the original size.
1 Like
Nitpick, but it’s very confusing. Since there is a beta 1.9.0.2b, and if code from that beta has been included in the new release, then the release number should be HIGHER than the beta, not lower, meaning this release should be 1.9.1 at least, and the current beta should close in lieu of a new one with a higher number. This is how beta/release numbering is supposed to work.
If the code being tested in beta 1.9.0.2b is NOT in the 1.9.0 release, as the numbering indicates, then my post is moot, but in the past this has not been handled correctly, so we don’t have a real basis to go on.
So, is beta 1.9.0.2b over, and is the code from it included in the 1.9.0 release, or …is it not?
1 Like
Strangely enough, I did a test earlier where I applied Artemis LQ v10 three times to the same video and it didn’t do much past the second pass. It seems to get it first time and there’s virtually no improvements with subsequent passes. I haven’t tried upscaling then downscaling but I’d rather not as I’m trying to piece together a quick and accurate workflow for professional work. If I can get to 90-95% of where I want to be with one pass at the scale I need, then I’ll be happy.
1 Like
interesting, since you’re in the beta program can you tell if some model in this release has a different version to the Beta? Maybe they might also be a different version without naming them a new number like it has happened recently with I think it was Gaia? I would suspect the file-sizes of the model-files might give a hint if that’s the case.
1 Like
Depends on the GPU, I would say. 
If the marked situation wasn’t so dire I would already own a RTX3080 but I guess this will take at least 2 more month till things calm down…
1 Like
I want to not just be quick, but as quick as possible. Running the card to its limits and reducing linear steps to a bare minimum. That’s why I’m interested in seeing the Dione model mature as I want to drop QTGMC and even pull back from using Neat Video for chroma and temporal noise removal as they both represent one step and I want to get down to just a few clicks.
My current workflow is the following for purely interlaced videos. Each step can’t be avoided or ignored -
- QTGMC
- Neat Video chroma noise removal
- Neat Video temporal noise reduction (can’t be combined with luma noise removal)
- VEAI
- Resizing VEAI output (because VEAI doesn’t allow for non-uniform scaling)
- Vegas
I want that workflow to be the following -
- VEAI
- Vegas
3 Likes
Mmmhhh I get everything but the Neat Video part. doesn’t Artemis handle also that? Sorry for asking if this is actually a dumb question but I have really no Idea. In general I always get the impression that most noise I encounter feel kinda natural to me at least for example old films and tv recordings, so is it just your special use-case or a general rule of thumb?
Will continue improving. Artemis v10 moves too much to the motion stable side vs the details generation.
2 Likes
If you are video is high iso/high noise, Artemis-LQ or MQ can denoise and upscaling at the same time. No need to run through any other denoise program.
Sorry about it. This release just focused on deinterlacing. There is a workaround involving editing registry if you wan to try.
Thanks for the feedback, The reason for a separate digital and analog videos is because analog has some specific problem such as color sub-carrier processing that yet different artifacts. We will see if we can eventually combine them and still product good result. Meanwhile, use TV if the video looks soft and DV if the video looks high quality.
Oh cool thanks for the answer. Could you maybe also answer some of my earlier questions?
BTW. one thing that would be a nice addition, since VOB´s are now supported, to add them in sequence. With some DVDs I always had a odd jump in between different files, that I could only solve by exporting the entire stream as one file.
It’s more for me really. I know EXACTLY what to expect when I use Neat Video as I’ve been using it for years. So I know it can do temporal noise reduction great and chroma noise removal without ANY loss of detail, even at the grain level. So I really should do a side by side of Artemis and Neat Video. I suspect I’ll be perfectly happy with Artemis in terms of temporal noise reduction. As for chroma, much the same, just need to do the testing.
Most noise is natural, whether it be the image sensor or on film transferred to videotape but when it comes to noise and artifacts introduced through DVD compression? Another story entirely as that’s wholly unnatural and needs beaten in other ways.
If your footage is out of camera, or “pure” interlaced video (not though some post editing), I am confident the Dione will be better than anything out there. However, most footage has gone though some editing and therefore not “pure” interlaced. QTGMC is more robust.
You’re probably right as the footage I’m working with is from mastered studio-quality VHS tapes and not straight from a camera. The DV tests I did with Dione were great and those were straight from a 25 year old DV tape.
We are not with Dione. But we are aware of the good properties of QTGMC and will steal good ideas from it going forward
Does this mean DVDs are already edited to much? And if so is this something Dione will be trained to better handle in the future or this simply to much variance?
Have you heard of VOBMerge? I use it whenever I forget to process VOBs as one complete file.
1 Like
I’ve given up on the model versions. You’d think they would not be updated without a change in model number (they broke that at least once), but for the sake of my question, I’m only referring to the software version itself.
