A bunch of suggestions, ideas and questions


First, i realy like to thank the hard working group behind any Topaz Products.

I realy like most of the new stuff that was added since AI 2.6.4 (which i still
use as my Main Topaz Tool, for some reasons. I just play around at the moment
with AI 5 and test out, what it can do better etc)

The AI Models come a long way since 2.6.4

Now I got a bunch of suggestions and ideas, and dont want to create one topic per
suggestion and idea (that’s spam/flooding).

I know upvoting per click makes it harder, and is probably not realy possible.

  1. Possibility to divide the FPS from the Video (simple way, without changing the Frame Time Speed, so that Vid & Audio is still in sync) as last step, after Doubling it first with something like Chronos.

Why ? I like what Chronos and Themis Motion Deblur can do, it will deblend and find/create good Frames, in a Video that was deinterlaced with blend (so there is a
lot of ghosting). But after that, i got a Video that looks to smooth/weird in motion.

Maybe it’s because i grow up with 25/30 FPS, and just restore Pron (99% is 4:3 stuff)
but this overly smooth motion is just to smooth, it’s realy distracting.

Any ant size motion from the cam looks like strong shaking, it’s like a cam shake catalysator.

The frames themselves look good, the blend ghosting is gone, but it’s all to shacky, people (and parts from them ^^) move unrealistic, weird speed, soap opera effect like.

  1. Less wasted room between Gadget Elements (all visible Gadgets: i call them so, because the coding language that i use for 20 years call them so)
    Stringfields/Inputfields, Bars, Toggles, Buttons etc etc all the Gui stuff.

99% from all people out there use widescreen displays, so Vertical Size is a bit
limited, and even more important, compared to people like me with 5:4 Screens.

a) The Room between “Fix Compression” and the next Slider.

It looks like the offset between sliders is 20+ Pixel. So many wasted H Space.
Even when using the fast way by dragging the gui elements on a 5px Grid together,
20+ px is extreme.

b) The Menu under the Title Bar.

If the Gadet (which slide the settings on the right side out, so that i can finaly see the complette seek bar with my 1280x1024 Screen)
is placed at a different position, there should be no need, that the Menu is always visible.

Make it so that it can be toggled (quick toggle with Alt Key) like in Firefox etc

c) The Seek Bar Area.

The Bar itself is around 20px, but the Vertical Line (position indicator), is around 50.

The Dark area behind the Seek Bar, is at least 150px high.

For what ? The Vertical Indicator, got 15-20 wasted px to the Top, there is a ~20px gap between the seek bar, and the scroll bar, just because the Vertical Indicator Line
is to high.

There is many Vertical room that can be squeezed, without loosing overview.

  1. Bring back the String Input Fields (so that people directly can type in a number, and dont have to waste time with the mouse), next to the Sliders (“Fix Compression”, “Improve Detail” etc)


Set the associated slider that was clicked, as the active gadget for cursor key events.

So if somebody click at “Fix Compression” then the Arrow Left/Right Key, should change the value in “Fix Compression” (and it must be visible, that the slider
area is now active. What’s up with that ? it’s never visible which Gadget is the active one. Even Flat Design Windows Versions, like Win8+ will always drag a Border around a active Slider)

What was the reason to remove the String Input Fields ?

Has it something to do, with the weird behavior in 2.6.4, were
“Fix Compression” would jump back from 59 to 58 (or 58 to 57)
for no logical reason ?

Maybe that’s still a problem, and unprecious slider jumping (very tiny slow movement, bang, value changed by 3) should hide that ?

  1. Bring back the Float num Grain gradiations from 2.6.4.

The Integer only Grain Size & Strenght now are to unprecious.

It’s even to much grain or to less, i can’t find a acceptable amount and size at the same time.

Stuff like 0.8 and 2.2 was great.

Especialy if i used Avisynth+LimitedSharpen, afterwards, to create a griddy look (out of the fine grain, that was created from Topaz 2.6.4 before), witch resulted in fake details. Which is awesome if the Source, was very undetailed/flat.

  1. Add a Tab (Toggle View) were people can drop, different parts, from the same
    source, with different settings, and join them together for a Final Output.

It is possible :slight_smile: Even i can do that with my Beta Tool.

I use Win 10, but i switched to Classic Design, because i hate Flat Design, and especialy the 1px thin Window borders.

Now to the Tool.

Any selected Line/Cut is a Part from the Main File/s.

I can join them (i let VirtualDub2 do the Job for me) to one single output file, like i want (i can preview any Cut with VirtualDub2 like i want: I click the line and press Enter. That’s it, or, i add a Join Line, select the Cuts that i want, use the right click menu, and Click on Preview. Non existend Cuts will be created on the way)

Any Line/Cut were is Red in a Tab, is excluded from Join Operations, even when selected (or marked).

Any Cut/Line got his own ID, Color Marker (Orange = Bad Original Form → PreProcess with AviSynth and use the Output for Topaz, Yellow = after Topaz now 4:2:2 → make 4:2:0 and import it in the right folder for the next step, Green = Make the Final process with Avisynth)

Any Line/Cut got his own Prefix, Tags, etc etc etc

I can assign the output (even if it doesnt exist at the moment) from one cut, as input to one or multiple other cuts below, with 2 clicks.

This is how i do it: Cramp as much as possible in a list, use Tabs “Main,Log,Filter Editor” dont let cripple your possiblities by Windows8+ Flat Design and Burger Menus.

Use unorthodox ways for pop up menus (the “Set INPUT -to selected” Button will pop
up a Right Click menu) and use a lot of them (the Right Click Menu for a Cut got 120 Entrys at the Moment)

  1. Use a Cache File (caching is blazing fast these days, even a mechanical HD can throw out 160MB/s+ Read Rate) for MPEG 1&2 and VOB Files, like FCCHandlers MPEG2 PlugIn for VirtualDub.

The reason why MPEG 1&2/VOB seeking in VideoAI 5 is so bad (jump around keyframes when using cursor keys, slow 1 frame forward backward, timeline fails to react anymore, Deinterlacing a Part from the Video and the Output will be a totaly different one etc )

It tries to read the Video Data like a normal Player.

Jump to Keyframe, if something between was clicked, jump back to the previous keyframe, or decode the last keyframe (slow !)

Another Advance of using a Cache File and a dedicated MPEG 1&2 Read is that
it will give precious and valid Data.

and it will determine the real lenght of a MPEG 1 & 2 File inside a Vob File.

Sometimes the lenght of a VOB File that is not the first VOB File (example:
VTS_04_2.VOB), will be wrong, and seeking nearly impossible.

  1. Let the People import multiple VOB Files (with same specs)
    as one continous stream of Video, and use MPEG 1 & 2 caching.

So nobody must rip first the DVD to one big VOB File (will destroy seeking), and extract (so no Re EnCoding) the MPEG 2 File out of the VOB File (then seeking works again), with something like Womble MPEG VCR (which doesnt work anymore with Windows 8+)

  1. Possibility for the Preview to show Different Aspect Ratios (and with different Resizers)

Here a example from VirtualDub2

  1. Possiblity to play the Preview a little bit slower as the real output.

I realy liked that behavior (in 2.6.4), because then it was easier to spot slight ringing or blocking, that is not a problem now, but maybe if watched on a bigger screen.

  1. Possiblity to Open a created Preview, with different “Players”

My Main Player is MPC-HC, but if i want to check out a result, and compare
different outputs i use VirtualDub2, and click around between them.

  1. Possibility to jump to “Mark Out & In” by mouse Click.

  2. Possbility to jump (by click) to the last processed frame from the active Preview.

  3. Add the Version of a Model next to the Name, like it was in 2.6.4

So that people like me, that use different models for different parts, know which model for which part was used.

So that i can do the same job under the same conditions in lets say 5 years again.

  1. Color Coding (Orange = Ok) a Preview, Tagging a Preview

  2. Add “Remove all groups except this” Option in the Right Click Menu from the Previews.

  3. Possibility to handle Luma and Chroma Noise as 2 different Types of noise.

By adding 2 separate Sliders for Chroma & Luma noise.

Because Topaz Video Ai has still the problem that it eats up subtle colors. (got a 10Bit Lut Monitor, i can spot that)

Now i got 4 Questions.

Q1) What type of Interpolation does Topaz use when displaying “Before” (at any Size) and “After” if displayed <> 100% ?

Q2) Do i guess right that Grain in Topaz 3+ behaves different ?

In 2.6.4 if i used lets say.
Artemis Medium without grain, it looked realy flat and to smooth afterwards.

Then if i used something like 0.8, 3.8 at 720x480, the output got way more detail and original colors left, and (i guess) Topaz 2.6.4 avoided grain around edges (which make sense)

Q3) Does “Recover Detail” overlay in, parts from the unprocessed frame, or from
a frame that was cleaned (Ringing & Blocking) up first ?

Q4) Does Proteus still work on 1 Frame base ? or is it now Motion based ? (like Artemis)

Sorry for the Text Wall ^^ and my typos ^^ english is not my native language.