It is the only way Topaz knows the aspect ratio.
Vegas has the best recognition of ratio that I have found.
Why not have option to de-noise and process in SAR.
Then another option to manually reduce height to match DAR.
Reading some history of program, it could certainly be done with tiff.
Hmmm. A quick read tells me that I’ve never run into that, but I can see that you have to consider that it may be the case for some of your users with cinematic video. I’m wondering what you do in cases like mine, which have square pixels, and the w:h is the correct aspect, but the meta DAR is wrong - how are you doing the squeezing to come up with the (incorrect in this case) output width?
I noticed under the Theia model, in the “Restore Detail” setting, a higher number setting (100) actually produces LESS details than a lower number setting (0).
I think it’s worked that way since the first version which includes Theia models. Nobody from Topaz has confirmed or explained it yet, so it’s unclear if it’s a bug or if the documentation is incorrect. (The documentation in general leaves a lot of room for improvement, but that’s a whole other topic.)
Just to be clear, I use a high setting and asked @mikaljan for a comment, who uses a setting of 80.
This could come down to the quality of video.
Some people want to convert SD to HD.
Others are only using this software for restorative purposes.
It could come down to the history of video; previously downscaled/compressed or just low res.
For me, 1.6.1 is a pretty solid release as it fixed a couple of issues I was having. However, I noticed something new. Let’s say you select a segment of a video and begin to process, sometimes it will “choke” at the beginning and skip a few frames. Specifically, it will process the first frame, then get stuck and resume processing a few frames later. In the resulting mp4, there will indeed be a few frames missing, between frame 1 and frame 8 for instance.
I’ve had something similar, where the video chokes momentarily, but I’ve not been able to repeat it reliably.
In my case it happens near the end (within the last second of the video usually) and I’m left with a video that has a frame repeated a few times near the end, which is quite noticeable with smooth motion. It makes no difference whether the video has sound or not.
Agree, it seems to be random.
Is it good to export a video in mp4 with a bitrate of something like 100 Mbps? Or is it better to export in another format or with other settings?
Thank you!
Hi,
I noticed that a lower setting with “Restore Detail” causes an output that is less “processed”, leaves more noise, blocking… etc. and increasing the “Restore detail” setting cause more processing done on the video, deblocking, smoothing…etc.,
This can cause a little confusion but I figured out what it’s really doing to the video, and I discovered that a “Restore Detail” setting around 80 does a good job smoothing the color banding and noise while retaining edges and details.
“Reduce Noise” setting is more like changing the threshold of what is considered as noise, the effect is quite strong for lower resolution sources. For the video I tested on, it was a 720x480i video with 4:3 aspect ratio or SAR of 1.3333. It is then preprocessed via Vapoursynth R52 with CNR2, FDR, QTGMC, Deblock, FineDehalo then Resized/Resampled with Gaussian Filters into 960x720p. Then this 720p video is used with VEAI 1.6.1 Theia v2.0, with AI resizing it with 2x resulting (1920x1440p). With such input video size of 720p, simply using only “Sharpen” and “Restore Detail” results in excellent processed video. Leaving “Reduce Noise” at zero prevents removing important details such as skin textures and fine hair on the skin.
hope that helps!
I would add that my footage of half your resolution, losses detail with reduce noise above 7.
Detail is also lost in my specific video, when increasing Sharpening beyond 55.
When I say losses detail, it is a loss of detail or reduction in removal of artefacts.
Thank you for posting this, very helpful. I think “restore detail” is a misnomer, perhaps “reduce artifacts” would be more accurate.
Yeah, I think that description suits it better
No matter the settings THEIA really lacks texture details. Topaz should more focus on the GAIA high quality models.
Anyone?
I always suggest using image sequences when possible. They are much more reliable and often produce better quality. I personally like doing TIFF because I have lots of HD space, but you can do PNG or JPG if you need to save space.
As Matt has said. I was typing so slow he bet me with an answer.
Thank you very much! I’ll consider a JPG lossless export and then reassembly it. Could you tell me a software to create an h.264 video from a sequence of images? Free would be better but I can have a look at paid ones if it’s necessary.
I hear FFmpeg mentioned, but have not used it?
Just mention that JPG is not lossless, but might suit your needs at maximum quality.
png would be better, or tiff 8bit.
Your video should have fixed fps, so video processes correctly and audio remains in sync once re-imported.
Well, I never used FFmpeg and I don’t know how to use it. I’ll look for some information on it!
Thank you very much!
It’s different from Gaia! why would you want all the models that does the same thing? I find Theia to be excellect and does a superb job for my purposes. It depends what suits you!
