Video Enhance AI v2.2.0

Theia is still incredibly broken, I loved that model! I’m using an RTX 2070 Super, and when switching to CPU, the artifacts go away. I am running 466.27 game ready driver.

Switching to studio drivers fixed the issue, however, I think fixing a game driver bug would be a good idea.

1 Like

I have a GTX 1660 graphic card. If I would change it by a e.g. RTX 2070: would be rendering time advantage in comparisson to the 1660?
Or in other words: what time it will take to render a 10 minutes video using GAIA with an RTX 2070?
With my GTX 1660 it takes around 26-28 hours for a 10 minutes video using GAIA only deblocking/improvment (100%).

Which compression factor do you use for MP4 264? By default it is set to value 14. But then the videofile is 2-3 times bigger than the original.
Would be e.g. 17-18 to get very good quality without losses?

The lowest should be CRF 15. It’s the best compression value for quality and file size.

The improvement should be 100% to 200% faster than GTX 1660.

For me, studio driver is better than game ready driver. Some tests and benchmarks were done on the internet.

I already use the studio driver since years.

Can someone please link me or explain here what does “h264 compression factor 20” mean. I come from DaVinci resolve and Premier Pro and After Effects world and I have been working with video for a long time and I have no clue what that seemingly arbitrary number means.

I would expect ability to choose my Bit rate and size calculated based on that setting, like in virtually all the programs. Why is Topaz using number that does not really tell us anything useful. At least if there was a proper guide as to what does the number “h264 compression factor 20” mean. Is there? Can someone explain or link me to it , please. Thank you.

Hi,

My understanding is that VEAI uses FFMPEG (hope that’s it).
The best is to read the FFMPEG guide: H.264 Video Encoding Guide
https://trac.ffmpeg.org/wiki/Encode/H.264

Hope this help.

Thanks for the reply, but Constant Rate Factor (CRF) mentioned in the FFMPEG guide, is not the same as compression factor mentioned in the Topaz interface. There is no reference I can find or explanation.

There is a whole slue of terminology mentioned in most programs. Everything from bit rate, bit depth, color space etc, but not compression factor, whatever that means. There is not even unit of measurement. Is it 20x factor or 20% of something? And of what? I’m sad to say, the user interface design in that instance is amateur. There is not even “?” little tool-tip indicator for it, even if there is one for “keep audio” option. Strange. Its like four different developers worked on it, and no one talk to each other. Where is consistency in design language?

Thank you for your link, but I remain with more questions than answers, sadly. I hope someone from Topaz team can explain it, and they can rework the interface for future releases so it makes sense and it is not only consistent within the Topaz interface and applications, but also consistent in regards to standards used by most applications already.

New UI is stuck on Artemis Aliased and Moire v9. Even if I click on a different model, and it seems to be selected, the model that is previewed or used is still the Artemis Aliased one. 16" MacBook Pro, VEAI version 2.2.0

More: Clicking a Theia model does not show the Theia settings.

I do hope Topaz will clarify, but I think compression factor 20 and CRF 20 are one and the same, as Topaz uses ffmpeg for video output. All I know is a value of 0 means lossless output but very large file, and 20 means much smaller file but compression artifacts will be visible to some extent. At a value of 15, it’s a good compromise, in my opinion. At the maximum value of 35, the quality would be bad.

Hi chuck, can you PM me your application logs? Thanks!

Slight confusion. The three betas had updated models, this release does not, i take it that means the model updates still being worked on and this is mostly UI and bug fixes?

1 Like

“I do hope Topaz will clarify, but I think compression factor 20 and CRF 20 are one and the same, as Topaz uses ffmpeg for video output. All I know is a value of 0 means lossless output but very large file, and 20 means much smaller file but compression artifacts will be visible to some extent. At a value of 15, it’s a good compromise, in my opinion. At the maximum value of 35, the quality would be bad.”

But that is like asking people how do you cook eggs, and the guy say, cook it until you feel it its ok. Than you will know. lol

We need well established numbers that we can use as targets so we can not work half blind. What is the bit rate and what is the file size based on bit rate. Or any other method of compression we are using needs to be clearly labeled and listed and allow users to change, otherwise its working half blind if not totally blind. Besides, ffmpeg as powerful as it is, it is also notoriously not user friendly. Which is really strange to use even less user friendly, arbitrary numbers while at the same time offering large thumbnails and examples an tutorials in the program for AI models. But for well established methods of compression it hides any useful info. [visibly confused].

Any news on the 3:2 pulldown remover ?
And if the pulldown remover can detect if a timeline is exported with clips with different startpoints for the pulldown ?

We kicked back the updated models due to quality issues reported in the beta. We ran a poll in the Facebook community and most people said it was a quality degradation compared to the last, so we threw it out.

We don’t have any plans on updating Artemis / Dione in the near term after this release, but instead are focusing our efforts on new models (FPS conversion, and a model with more user control).

2 Likes

On some cases Artemis V13 was giving good results.

For those who want add HDR to videos ( for anime, movies or anything ) use this software called Shutter Encoder www.shutterencoder.com , also for people who are interested check this comparisons grading HDR example using Shutter Encoder - YouTube

I use Video Enhance AI + Shutter Encoder for HDR