Video Enhance AI v2.3.0

the reason its not matching is because for some reason chronos is stretching the footage, im glad to see im not the only one who experienced this issue. thanks for sharing your experience. I hope a developer is looking at this feed so that they can report there is an issue with chronos. Also this same bug happens to me on the dtd models that interlace that doubles the frame, it never keeps the video length the same. Its supposed to double the frames but it shouldnt be making the video longer. Please devs look into these two issues.

1 Like

I havenā€™t checked the length of my video after processing but you are right. When I inserted the original sound it ended long before the video, so the video must have been stretched!

1 Like

At one time it didnā€™t, then it did, now it still does (if you choose clean installation), but it doesnā€™t do it well.

Flowframes has support for this. But itā€™s a bit prone to artefacting. Works well when it works however.

Is it better to first deinterlace with interlaced tv and then proteus?

Has anyone noticed how Videos processed with 6 channel aac (maybe all formats: ac3, dts, etc???) audio in MP4 always come out in 2 channel only??

1 Like

Both Artemis and Proteus models have

  "interlacedFrames": 0,

at the top of the JSON. (C:\Users\[USERNAME]\AppData\Roaming\Topaz Labs LLC\Video Enhance AI\models\proteus-b-1.json)

I wonder what would happen if it was changed from 0 to 1 and you passed interlaced video through it?

Really I donā€™t knowā€¦ I havenā€™t triedā€¦ it might not do anything, this might be a setting that would be used in the future maybe? Or maybe it does something now? Only one way to find out!

You must considere that actually the audio part of VEAI is there just forā€¦ Art ? lol.

so yes, actually, the audio part is transcoded in stereo AAC, and doesnā€™t support any great bitrate or multichannel or anything. iā€™m sure itā€™s on the list of a future release, but for now, itā€™s not.

Can you send me your logs?

2021-07-06-18-25-38-Main.tzlog (33.9 KB)

Here it is, first I load a video, then set auto with Proteus and it works fine then I remove the video and load another one and as soon as I click again the auto button the program crashes without even any error message.

the Topaz team told me that an improvement on the audio quality will happen in the may 2021 as far as I remember but it was a bit delayed unfortunately. :neutral_face:

I just tested it out. It makes the program really unstable and I wasnā€™t able to finish processing even a short video without it crashing.

yes, they told me the smae, so it will there soon ;).

I believe that is because of some set back in Artemis v13 and Dione v4

The audio overhaul was tentatively scheduled to be part of the June release. It wasnā€™t. I have no idea where it stands and despite any possible future company communications about itā€¦Iā€™ll believe it when I see it at this point.

Yep, over a year ago. Just donā€™t do audio in Topaz. End of story.

I just realized that VEAI when importing the videos, it donā€™t do it completely, some final frames are missing. Even if you upload a video already edited by VEAI, it cuts other final frames again ā€¦ they are few but it always deletes them. I have checked other videos that I already edit with VEAI and it has happened in all of them, so it is not only in this version 2.3.0, at least in my case.

Well this sucks!

With the beta I can import a series of png files into VEAI and encode a 15fps video at 30-120fps with the speed at 100%, yet Iā€™m stuck with only 30fps in the general release!

Why is that, and can you guys please restore the functionality?

This is why I never uninstall the betas alongside the general releaseā€¦

For good quality 720p source video upscaled to 1080p, Proteus doesnā€™t stand up to Gaia HQ in terms of generating low level, fine details, no matter which settings are used. Gaia HQ results tend to look more realistic.

Hopefully, Proteus v2 will close the gap further and v3 will be better than Gaia HQ.

3 Likes

The Gaia model files are much larger than Proteus / Artemis / etc. Larger model files need more processing time.