Tip: Three step tuning method, to recover bad quality videos

  1. Manual mode or relative Auto mode?
  2. What about the rest of the parameters, what value you set them?
  1. Manual mode or relative Auto mode?
  2. What about the rest of the parameters, what value you set them?
1 Like

relative auto mode for each of the models, and other values at 0

So basically the rest are on Auto mode
 That is why it is always important to say if you use Manual Mode or relative Auto mode, as value of “0” in both modes mean different things.

Yes. Sometimes when there is no noise in the video I use Prob4 relative to auto with improve detail 100%

1 Like

Manual.

I have.
It takes quite a bit of time and doesn’t really offer much of an improvement.

For the moment I’ve remained at 4.2.0

1 Like

Yes first pass Proteus-denoising (manual mode), I do this too in some cases (without upscale), but often at low values (5-10) to keep details.

The denoise function from Proteus is harder than the one from Iris and I like it. Ok we have NYX3 but this removes the noise from the smallest gap, this can result in very hard contours, Proteus does it not at this depth, more overall.

But if source has low noise/fog, it’s better skiping this step and you do denoise within main pass and loose less image information.

Recovering Details at the first pass instead at the last, results in harder object contours. This could be good or bad, pending on source and result that you want to get.

Is DS9 film on DVD interlaced? I think not, give try set progressive in Topaz. You see interlaced or not by fast movements, if there are no frayed contours it’s not interlaced.

The whole interlaced stuff is a curse, since CRT screens could only display interlaced DVD players produce fields when playing back.

But that doesn’t mean the source material on the DVD has these image offsets, most of it was filmed in progressive, but not always

Thanks for this great thread, really informative. Relatively new to this, so can I ask would this potentially apply to animation too or just live action?

Currently trying to upscale Spider-Man the Animated Series (1994) on DVD and getting some mixed results using Artemis. I’m getting some mixed results, close ups look great, but wider shots have a bad Haloing effect.

I’d go with GAIAI CG for animated stuff, Artemis does too much denoising and over-sharpening which is causing the halo. This is the least aggressive model for upscales with only minor sharpening applied. The Other option would be going with Proteus which would allow you to tweak it more to your liking but chances are you’ll still get mixed results. The dehalo options just blur so I wouldn’t use any of those personally. Also if it’s from DVD it’s probably interlaced so you may want to use the Dione model before you upscale.

Things changes
 now often it get best results by using model “Rhea” in auto mode. But Rhea generates less shiny results than you can do with Iris or Proteus. But here is a way you get some of the “good” from the other model:

  1. Enhance with Rhea, in Auto mode

  2. Make a second enhance using Iris, sharp but smooth, fewer contours, using a bit negative values of “Antialias-/Deblur” maybe a little bit “fix compression” The goal is making a crystalized shiny result a “fluid” one, it may also be a bit frayed, this is what we want.

  3. Overlay Iris and Rhea vids in our video cutter software and reduce opacity of the overlying track, for example to 50%. Result: Both video tracks merge into one. Adjust brighstness/contrast/gamma as you wish an export.

  4. Optional: Reimport the merged video in Topaz and “polish” with Artemis HQ or Strong Halo

2 Likes

When was this Rhea model introduced? I cannot find it.

Rhea was introduced about a month ago in v5.2

1 Like

Have you tried using the new Focux Fix module, if so with which enhancement model do you recommend.

I’m using the Focus Fix to quite some success recently.


Voyager - Scorpion - Chakotay confronting Seven of Nine

1 Like

I realise this was posted almost a year ago but people will still see this when searching for this problem. I too found the same issue - after I began using Hybrid QTGMC for de-interlacing. Suddenly all my input videos were showing impossibly high FPS. The solution - for me at least - was to export the original source material from Lossless Cut in a MOV container as opposed to an MKV container.

1 Like

or this way yeah, the main thing is that it works :slightly_smiling_face:

something else, tip: for audio fps/speed conversion check out “clever FFmpeg GUI” I love this tool. Directly drag you’re video in it, then click on “main” → “audio stream” → “goto encoder” and under target fps enter a value, then select “length only” or “lenght & pitch” choose codec and click “convert” :yum:

Sorry but where are the settings “Recover Original Detail“ and “Recover Details”?

Are you actually referring to “Recover Original Detail“= “Recover detail“ and “Recover Details”=“Improve detail”?

The above “Parameters” one was earlier called “Rec. Orig. Details” and then Topaz renamed it. You can still see the old name in the video container description with mediaInfo.

But this threat is outdated now. I spend a lot of time finding out what works and what not and have a new workflow now, it gives me some stunning results, it’s really amazing. I do combine Iris with Rhea layering and I found now the perfect finisher both “melts” into one. But it is time consuming and hard to describe doing the right thing pending on source video. I open a new thread then later, when I have time to wrote the guide.

1 Like