Video Enhance AI v2.1.1

No problem, will deliver the message to him :slight_smile:
DVX100… that may explain why it’s a 3:2 pulldown on the dv footage.
He probably filmed in 24p mode.

I am not sure about the 24p mode. This is what I found on a site about the 2017 remaster for Bluray:

The original DV 30 Mbps 29.97fps media was converted to a new ProRes 422 HQ 220 Mbps source at 24 fps for more cinematic motion and for more control and manipulation of the picture during an all new color correction process, with each shot meticulously noise reduced, sharpened, and detail enhanced.
The Man from Earth Restoration Comparison and Blu-ray

If he really had filmed in 24p then why would he had done a conversion?

EDIT: Found this about the DVX100A - it has two modes:

24p Standard utilizes the same 2:3 or 3:2 pulldown cadence long used to transfer 24fps film to NTSC video. The first 24p frame is written to two fields of 60i video, the next is written to three, the next to two, and the next to three again

24p Advanced uses a syncopated 2:3:3:2 pulldown cadence to stuff 24 frames into 60 fields. It’s ever so slightly different in its playback; the standard cadence of 2:3:2:3:2:3:2:3… evenly intersperses the “short” and “long” frames, while 2:3:3:2:2:3:3:2… lumps two “short” frames together followed by two “long” frames. The difference is subtle, but can be seen on smooth pans or on regular in-frame motion, like the passage of a train. In my experience so far, about half the people looking at a 24p Advanced clip can see that the motion is a bit different, and half cannot.

But 24p Advanced isn’t intended for making the 60i video look like film; it’s designed to allow the best possible recovery of the original 24 frames. You’ll note that all four original frames can be recovered from self-contained 60i frames; the green frame in 60i now contains the “extra” B and C fields and can be discarded, since all the information for B is contained in the yellow frame, and all the information for C is in the magenta frame.

Extracting a true 24p clip from a 60i recording simply requires copying the raw data for the red, yellow, magenta, and blue frames into a new 24p file, skipping the green frame altogether. No decompression or recompression is required, and all recovered 24p frames retain first-generation quality. No clipping or other loss is incurred; you are still working with all your frames in their first-generation glory in 24p.

please please please fix the jpeg output… especially during fade ins and outs there is horrific jpeg quality.

please compare:

Ok. So i was right saying to him that the footage i got was not “pure” dv footage.

Here is a comparison of the different models :

Lower VRAM usage in the Preferences. Reduce machine load.

Roll back to the 461.92 game drivers/the latest studio drivers.

Is it just me who thinks previous AI models had better results?

The new AI models, v10, v11 and v12 seem to be overdoing what v9 was doing almost perfectly.

Does anyone else think the same?

1 Like

I get these “lower your ram usage” message too in two of the de-interlacing modes no matter if it is set to the lowest possible already!

How many instances are you runing?

Exactly one. :slight_smile:

What’s your GPU?

Using an Nvidia GeForce RTX 2070 Super Gaming X 8 GB.

Do you overclock your GPU?

I made experiments with and without overclocking. The 1.7.0 ran with overclocking perfectly 17 hours non-stop on many files, so I assume when it gets the error on standard clock or even lowered clock something else is going wrong. :neutral_face:

Maybe you should use Studio Driver in Nvidia Geforce Experience

Running a GTX 1070 8GB with 465.89 Game Ready Drivers. “Reduce Machine Load” is also turned on, and I haven’t had any issues running VE AI.
image

Hi,

For these Nvidia GPUs, it seems that Topaz recommends Studio drivers. I found in an Nvidia article from November 2020 an article on the Nvidia site.

“November brings the release of a number of new app updates to boost your creativity including Blackmagic Design’s DaVinci Resolve 17 Beta Keyshot 10, Notch, Topaz Video Enhance AI, BorisFX Continuum 2021, Red Giant Trapcode Suite 16 and Magic Bullet Suite 14.”:

Incredible Video Upscaling From Topaz Video Enhance AI:

The latest software includes performance optimizations to reduce overhead, improve model load times, and reduce VRAM usage.

I’ll be tempted to find a version of the driver from November 2020 to see if it’s better and so on until the last one and compare which one works best.

Viktor, I am already using the studio drivers. But I am not using the Experience tool. I think that one is for gamers. :slight_smile:

Then uninstall it and reinstall Studio Driver from Nvida Geforce Experience. Also restart your PC then. While upscaling, don’t watch Youtube.

Hi,
How could I send you a file to check sth if you could be so kind?
Greetings,