Topaz Video AI v3.0.3

Hello Everyone!

A new version of Topaz Video AI is now available.

v3.0.3
Released November 8, 2022

Download: Windows | Mac

NVIDIA users: Driver version 522.25 or higher is required

Changes from v3.0.2:

  • Fixes exporting to NAS drive.
  • Fixes importing/exporting EXR image sequence.
  • Fixes stabilization when there is a single quote (') in the Temp/Export path.
  • Fixes right panel scrolling to the warning message.
  • Fixes Interlaced model copy-paste.
  • Image sequence can be output up to 16k.
  • Improved in-app download speed.
  • Misc. bug fixes and improvements.

Known Issues:

  • Quality mismatch between 2.6.4. and 3.0.x for some models.
  • Some users may experience reduced performance compared to 2.6.4
  • Preview experience needs to be improved
  • In-app updater may indicate 1 export is being processed although there isn’t any

File Submission Dropbox: Submit Files

5 Likes

My impression on this new version, still the problem of the preview during video encoding which does not scroll when encoding the video export. Still we can’t choose where to put the old models on this version 3 as a lot of people ask here, and it’s better to put the list of models back so that we can choose the model we want ourselves. wanna. the encoding speed is still very slow, this problem will have to be remedied quickly since the official release of this version 3. The artemis model is frankly not excellent. Proteus is the same, too bad the quality of these models are in losses. Between 3.0.2 and this new version I really don’t see any difference. but really none. The impression that nothing has really been settled or is it me who sees nothing? Are we writing in a vacuum on this forum? I’m asking because I don’t feel like we really take into account what we write to the TVEAI team… :hushed: it’s just my impression. here is my opinion on this version 3.0.3. I hope my opinion will be heard and taken into account :blush:

10 Likes

Correction: Preview must be working at all. I understand their reply as either a miss understanding on the importance of the preview or as a giving up having no clue how to fix it. :no_mouth:

1 Like

I understand your message but I don’t understand a thing, we always report the same problems, bugs, and others to them and we have the impression that nothing is taken into account behind. don’t take this the wrong way Team Topaz it’s not my goal. I respect your work, but for me I really don’t see any improvement since you created this new version. After that it’s just my opinion. Why in this version we can’t choose our own models? why can’t we put the old ones as it was in version 2.3.0? I don’t denigrate this version 3 but it’s really a return to square one and it’s really a mess. I don’t know what more to say. I hope you will really realize that there really is a problem since you wanted to create this version 3. I’m waiting to see the rest :wink::+1:

8 Likes

Preview still crashes. This app becomes impossible to use.
Don’t you test your updates???

4 Likes

Preview is fixed for me in v3.0.3. I can generate many previews. Pan and zoom during preview playback now work well with no lock-ups. The processing is slower than TVEAI 2.6.4, but the quality of Proteus output is better.

NVIDIA RTX A4000, Intel Xeon Silver 4216, Windows 11

6 Likes

i confirm that for the only one test i did tonight the preview is working here on 1050Ti Gtx, Intel i7-7700, W11 Beta Insider Preview. as i have on a second computer a A4500 RTX, it’s certainly the same as you. will check that in few hours.

1 Like

There are only a few million combinations of hardware and video codecs. It’s impossible for them to test for all of that. The best they can do is ask you to upload a sample of the crashing video to their dropbox (Use the link in the first post.) and let them know what hardware you are using that crashes.

2 Likes

I use a translator.
I’m glad to see that it now retains the model and settings after an update.

I ask the at the release of 3.0.2 could I use an older driver(517) instead of 522 for Nvidia cards. The answer was yes. The my point of the question was actually we get no explanation why the change of using a new driver. Was it for performance or quality?

Well I used 517 but 3.0.2 was still buggy so I quit using it and back to 2.6.4

I tried 3.0.3 and I also like many had it crash as soon as the preview started. I decided to try the latest Nvidia driver and game ready at that. Guess what the preview now works.

1 Like

I really hope that 2.6.4 can be surpassed in terms of speed and quality in the next few weeks.

6 Likes

I’m the only one who wants to review the list of models for version 3? does anyone want to see this again? nobody talks about it but me?

12 Likes

I remember they said they would work on the models after the official release and so far there is nothing on that side. Why?

4 Likes

I’ve noticed the gap and I’m glad you bring it up. I just can’t use version 3.0.x so I don’t experience the gap often. I can’t waste time trying 3.0.x until it is feature complete and stable. Tbh the video quality of version 2.6.4 is less than stellar–I didn’t find it lives up to the claims even for a good quality progressive DVD to HD upscale (far too many obvious artefacts from the upscaling process). However, it is much more functional than 3.0.x and it doesn’t crash.

2 Likes

I use an old version of VEAI, 1.9.0 and 2.3.0 because I can really use old models which were better than today’s :blush: and which really had no bugs. I don’t understand the TVEAI team, because we’re all complaining about this new version 3, it doesn’t boost them to come back to version 2 to continue it, I don’t understand this new team.

1 Like

Still no face enhancement ? You should really include something like codeformer face restoration in the next update. It’s high time.

6 Likes

Just tried clip for 2.5 hours processing - no crashes.
But still no .avs support…

1 Like

it really needs to solve the problem of encoding which is seriously very slow. And that they update the models and especially to put back the list of models so that we can choose ourselves as it was in version 2.3.0 and that we could put back the old models. Already it will be a superb progress for version 3 for my part

2 Likes

Same performance as before

only difference is: log file cannot be opened because it is used be an other process