Topaz Video AI v3.1.5

I really don’t think the problem you cite is from CRF encoding… but rather from some issue with the old program!

1 Like

The CRF encoding method has unfortunately been removed because it is only applicable to software encoders such as x264 and x265, which were not added because of a “licensing” problem of some kind-they are freeware software, I really don’t understand this limitation yet!

using open source code does not mean you can do anything with it. there are limitations related to the use for commercial purpose. If the licence allow it , it has some “consequencies”. as an example, one of them is that the software which use the “open source” software under licence is to put the software which use the source under the same licence, or to publish his whole code, etc… there are some other kind of limitations depending of the licence.

Because you add that yourself through an option in your personal preferences.

1 Like

Sigh. I just ran a 4 hour job and it came out with a watermark. The other job that was running at the same time came out fine. Never had a watermark issue before.

It would be nice to be warned when, using the CLI, if it’s watermarking, or in trial mode, or it failed to login.

2 Likes

I also spent a day carving the watermark out of a 2 hour movie.
I would like to see a solution to the license error, like making it so that FFMPEG cannot be started without some option.

1 Like

It doesn’t need to be so complicated. Although none of the existing encoders support CRF, most of them have similar solutions. The current graphics acceleration technology makes it several times faster than CPU encoding while maintaining a certain quality. Most users should not need it, of course, the premise is that you have compatible hardware.

TBH I use nvenc_hevc qp encoding and visually it has no quality loss at least to my eyes, and a good file size.

So x264 and x265 are not essential encoders, Video AI just lacks a parameter setting.

I don’t need them but many others have stated they wish they could have them. Due to licensing/GPL issues it will never happen though.

Yes, unless Topaz finds another similar encoder to replace it.

I just had Topaz Video AI crash on me, when it was just completing a video (3.1.5).

This is the first time it has happened to me. I was using the Artemis model on an SD video, upscaling to 720p.

Are others having the same issue?

1 Like

Great job for new version. I can see significant quality improvements for my home videos :slight_smile:

I want to throw in some ideas for potential future features:

  • Compare Previews in a side-by-side manor: It would be nice to compare the results in one single view side by side to see the differences. Switching Previews from the list takes a few seconds and is not that useful, when comparing results.

  • Output Video compression: For some reason the output video is quite big in my case (mp4 h264 high). Maybe you could implement a compressing feature to compress the output video for better storage size and web transfer.
    EDIT: Just saw, its already on the Roadmap. Great! Sorry.

  • I long whished feature from myself is an instant Frame-Preview of the current settings: If everytime I change some values I need to start preview calculations, it would take for ages to get the best result for the scene. It would be nice to click a button to preview the current frame based on the settings without rendering a whole sequence of frames. (I know this isnt possible technically yet, but it would be a very nice feature)

Again: You did a very good job in this things. I am very happy and recommend this software masterpiece to everybody, who is interested in this topic :heart:

Kind regards,
Launemax

2 Likes

The frame increase engine is still blurry.
The Themis engine gives excellent results for some materials.
The Themis engine works well with the stabilization engine.
However, do not use it with the Proteus Engine.
I have great expectations for future developments.

mage Sequence Import crashes Video AI 3.1.5

The moment i try to import a sequence the programs crashes… what can i do?I want to save time by not making a quicktime out of the renders-PNGs first.

The PNGs have alpha, could that be a problem?

Me to, but i managed to save the prores.

Indeed - I also had this problem. The final file appeared to be still writing out but the green tick was there.

Oh…and if anyone reads this, a pause function would be appreciated

It’s still failing for me when I try to preview with Chronos Fast (fails to download the model). Also fails with Apollo.
It’s previewed with Gaia okay (it downloaded it okay - one of the files anyway) and it’s downloaded what I think are model files starting with prob-v3 (proteus?).

Are Chronos and maybe Apollo files a lot bigger than files for Gaia? Could that be the reason it fails to download them? It spends ages in the “downloading mod…” section but doesn’t give any informative info on how much it’s downloaded (just shows the 3 dots after “downloading mod” moving). I have 12 Mbps download speed so they should be downloading reasonably fast.

Is there any alternative servers I can just download them all from so it all works?

I’m assuming Chronos etc. are currently failing due to timeouts due to it taking so long to try to download it (with maybe a lot of other people trying at the same time).

1 Like

Even if in version 3.x the preview can not be instant anymore it could even with a delay run continously by taking into account the processing speed. But don’t listen to me. As my former friend Martyprod already said, Imo with the few atari games he made feels too none-professional to him. While I greately appreciate his honest opinions to me I do not want to get any further reply by him anymore. Thanks. :neutral_face:

1 Like

Why does setting the output to 7680x4320 (4x upscale of 1080p) with export encoder set as H264 high (nvidia) (the only h264 option that it shows)
say "Error: For current encoder the output width should be in the range 145 to 4096 and height between 145 and 4096)?

Mp4 itself does allow 7680x4320. If there’s some limit with an nvidia hardware encoder then surely they should also have a software option that encodes to 7680x4320 as an h264 mp4.
eg. wikipedia shows level 6 of H264 and above all allowing 7680x4320 or higher

Video Enhance AI allowed 7680x4320 with “Mp4 - H264” (it didn’t show “nvidia” next to it) even though it gave a warning saying “outputs larger than 4K may not work with mp4” so why does Video AI have a lower limit? It’s supposed to be better than Video Enhance AI so surely it should be less limited in terms of resolution (including with mp4/h264) not more limited, especially since one of the things the app is promoted for is upscaling.