Topaz Video AI v3.0.4

To be honest, I’m still not a fan of the new GUI. But I find it really bad that you can no longer manage and use the AI ​​models offline. I also find the new procedure for the video preview much more cumbersome now. I don’t need multiple previews. If I changed a value and clicked preview again, it was fine. Now every time I have to delete the old preview, go back to normal mode and click preview again. I have a lot more that bothers me, but unfortunately my English is too bad for me to be able to describe what I mean properly.

Of course, the update also has positives. I think it’s great that you can now use several AI modes at the same time. Also that the audio track is now preserved in many more AI modes.

The bottom line is that I regret installing the update over the old version. Well, the software is being further developed, so I have hope that the offline AI models will come back at least as soon as possible.

1 Like

Thanks, I’m happy using ffmpeg from the command-line (indeed that’s how the video that went into TVAI was processed just beforehand) but thought/hoped tvai would/might do better work on the pans. FlowFrames is new to me, I’ll give it a go.
Thanks again for the comprehensive answer. (Oh and BTW Lanczos is generally my first choice.)

U open a command line window from topaz (main menu), then u copy your export command (main menu) and paste it into the command line (removing the nostats part will allow for progress display). Press enter to start.

When u click into the command line, execution will be paused. For resuming, press enter.

Thank you, it works

Again TVAI 3.0.4 Proteus creates strange patterns:

2 Likes

by curiosity, what’s the size of your video source ?

1920x1080 upscale to 4k Proteus auto

from what i know
1/ download model manager is scheduled to be implanted at a later date
2/ more way to display previews are planned as well (like multiple rendering on the same screen (like on the 2.6.4) and much better.
3/ previews are scheduled at some points to all be in sync so it will be easy to compare them.

“only” issue (lol) is that we don’t know when it will be implanted in the software.

I hope that at some points, one of the dev of the software will come here and maybe answer to some of our questions, with maybe a time frame about what and when it can be implanted.
you can reinstall the preview release anytime (2.6.4).

1 Like

i see you’re chaining all the models in one “go”. i would suggest to do them separatly. there is a very high chance that at some points in the 19 days, something can go wrong and all this will have been done for nothing.

your preview is not normal. when we look at the source, there are some patterns on the image. maybe it’s related to that. just do an export on a small excerpt of it where the preview fail,to see if the issue is there on the export part too. it’s maybe related to the fact you use all 3 filters at the same time. do you have this issue if you only use one model at a time ?

1 Like

there is a missing thumb down button on this forum, because you’re fully deserve it since you’re there. the only thing yeu’re able to do is to send trash comment to the dev team, the software, you’re blocked in the past, requesting the same things overn and over again, like blocked in a loop hole, stuff that have been answered to you many many times.

Don’t know what’s your problem, but this behaviour is not good for other people here and help in nothing.

1 Like

Av1 encodind is only for people owning a 4090 RTX card. it will not be implanted on 2.6.4 because NO more update of the 2.6.4 version will happen. it has been repeated a thousand of time. 2.6.4 is from the past. 3.0 is the future. even if the softwrae is actually not working like all people would except it, it will , at some point. i’m sure of that.

1 Like

As far as Proteus, I think it’s the way TVAI 3 handles the models that’s a bit different. I re-downloaded the latest Proteus models into 2.6.4 and there is no difference in results within 2.6.4 itself.

I see this problem on Photo Ai too sometimes on specific picture. maybe it’s related to the way the Model Ai handle some specific part of an image, like some specific “noise” or “parts” of a face. hope someone will answer about that, and a fix will be available at some point.
I saw this “problem” with other Model Ai generator from other software too, not only topaz product !

I have not decided yet if the Topaz team is wrong, but you make some good points. So far, Topaz is still ahead of all competitors in usability and overall quality.

1 Like

Stop spamming the forum with the same thing over and over. This is your last warning.

2 Likes

yes! :slight_smile:

1 Like

Some things that I really like in v3:

  1. Cropping the video size, easier than 2.6.4
  2. Processing up to 3 videos at the same time, while the UI is still live and not ‘frozen’ like in 2.6.4
  3. Copying original audio track to output file (no loss of audio quality)
  4. The ‘pause’ functionality is coming in the future (hopefully soon)

These things I believe it is not possible to add to 2.6.4 because of its technical limitations.

2 Likes

I haven’t had a lot of luck with 3.0.x releases…

Have some that ran for quite a while and resulted in some frames that went to grayish metal color and blocky.

I have not tried using the “Add noise” setting, will that help on this?

you are right, it is a singularity, normally Proteus Auto works fine when the input is “clean”.

I don’t know. Just give it a try?

1 Like