Is there a setting you can add to make the output B&W only? I get slightly pink lips on most faces on any setting I use. It would be nice if it colorized everything but lips only is a little distracting.
1 Like
I’ve had to go about this method also when scaling progessive video from the internet where the source was incorrectly processed from interlaced to progressive leaving some trace scan lines. The issue is, unless you select interlaced as the video type on the first pass, you can’t access Dione. This was a DVD upscale he mentioned, so his source is progressive. In the progressive vids I’ve scaled, I’ve gotten better results using Iris first and then a second pass with Dione, as Dione does become available as a second pass option. The only issue I have is, while Dione TV seems to do a good job getting rid of some of the scan line artifacts it introduces color banding.
1 Like
It works for me as I described it 
Hi, can the Topaz research team please prioritize low quality videos for Iris V3? Meanwhile, it would be nice if Iris V2 becomes more better in handling low quality videos in upcoming updates, as we wait for V3.
Iris V1 was a game changer. Please continue the good work, thank you 
7 Likes
I was using the beta 3.5.1.0.b on Ubuntu with 2 ARC A770s, almost seemed promising but there were too many bugs. I switched back to Windows 10 to try it…
ARC GPUs work but it’s weird. All GPUs selected causes the render to stall, both GPUs and CPU (5900x) go to 100% and it has to be canceled. When running 1 GPU selected, GPU-Z reports very low usage, like 10% but it runs fine, I don’t have a good grasp on how fast these should be but it seems fast enough. It’s definitely not using just the CPU because it’s way faster. This would probably run pretty good if it actually used both GPUs
It almost seems like Topaz is running models targeting an Intel iGPU for ARC. It was mentioned in the beta thread that you guys don’t test on ARC. Is there any way to get properly trained models for ARC? not sure what’s all involved but I’d train them myself if I could.
Please do not, one post does not speak for all users.
I’m perfectly happy with the models being refined for MQ inputs.
What are your settings here?
What model?
100% or Upscaling?
Auto or relative to auto or manual?
what is the value of add noise?
Have you changed the colors too? in the second image are better.
Bug - I exported my video and now it’s showing up in landscape mode.
How do I fix?
Hi, All you need is to upload to a file sharing site such as DropBox and share the link here. The other option is to raise a support request at the main website and Support will provide a upload link for you.
Many of us just want the Artemis type setup: one for LQ, one for MQ and one for HQ. Maybe add Proteus and Iris to that if we have the time and inciination to occasionally play around with sliders to raise quality from “improved” or “much improved” to “perfect”. And no bells whistles or frills that can be done as well or better with any old non linear editor that most of us, I suspect, already have.
If Topaz want to sell this to the “perfection” market, all good and well but say so - otherwise first things first - don’t forget the many of us who mainly want something quick and easy to get good improvements but who are not striving for perfection and who aren’t interested in spending hours tweaking half a dozen sliders to get that last 5%. Two distinct markets so either cater for both, or say which one they are aiming at.
2 Likes
If y’all want an improved LQ model then you should post it as an idea and vote for it.
2 Likes
Thanks for the reply, Don.
The support ticket link now forwards to a documents repository instead of a report form (either support.topazlabs.com or docs.topazlabs.com). There’s no obvious way of opening a support ticket. Historically, a dropbox link has always been provided here…so…
I uploaded to Anthony Martinez’s 3.4x file drop with Dropbox. I don’t have a file link as it doesn’t provide you one when you upload. The file is “sample.mkv” and is 8.8MB in size.
Dropbox is no longer free for me, my file server is undergoing a storage pool migration, and I don’t have the inclination to spend additional resources on this. The onus, frankly, shouldn’t be on the customer for this sort of thing anyway.
There is a blue Callout on every page, popups need to be allowed in your browser, where you can click and raise a support request. You can also use help@topazlabs.com
Just like your post doesnt speak for all users. I believe part of the vision of Topaz AI was to bring new life to LQ to Low-MQ and a good many of us are using it specifically for that purpose. So as part of that user base, which is likely half of us, Id like to see some priority given to LQ as well. Obviously you are happy as a MQ-HQ user and thats great for you. However, for us on the other end of the spectrum, where Iris V1 was a game changer for LQ, Iris V2 was not. Yet its still labeled as an LQ-MQ version. Its does well on high-MQ to HQ, but isnt great as V1 for LQ. If Topaz is going to label it as an LQ tool, then priority needs to be given to it to improve to that end, if not, then label it a MQ verison, like they do with Artemis, so the expectation is understood its not for LQ sources.
5 Likes
The issue we experience is quality in > garbage out.
As suggested, I’ve created an idea topic to get user input on giving priority to optimizing V3 for LQ or adding a separate LQ Iris model.
Vote away, group!
4 Likes
Nice idea.
(I removed the link, since you added it.)
Thanks, ForSerious! Also added the link to the post.
1 Like
-
hybrid (qtgmc deinterlacing, dehalo, range converion : pc>tv*, resize)
export in square pixel 768x576 (lancsoz) -
vai: artemis hight (recover original detail 95) + gaia hq (768x576)
export to ffv1 -
vai: iris v2 (relative to auto) (1536 x 1152)
export png 8 bit -
vegaspro (magicsharp, details+, dvfixer, color corrector, film grain) (prewiev 1280x720)
the only problem I have is because of irisv2 which gives me artifacts in certain places. I tried to do irisv2 + gaiahq, it’s fantastic, the image is even more natural and the aliasing effect completely disappears but the artifacts caused by iris v2 are amplified with gaia 
I think that gaia after iris can work miracles for everyone with this workflow because in my case, it does not remove details but in addition it removes the little residual noise that remains while smoothing the jagged edges well.
*range conversion is very useful for recovering details in dark areas.
3 Likes
