Although Ver3 slowed down the stand-alone work, when two videos were processed in parallel, each increased only 20% of the time.
Normally, it should take twice as long, but it only increased by 20%.
If multiple videos are to be processed, Ver3 is faster than Ver2.
I use Macbook Pro 14inch M1 Pro 16GB.
2 Likes
3.0 still looks like Beta with all this ugly fonts and buttons on strange places. Even my students make more professional-looking Java projects. But main problem is speedā¦ On my RTX 3090 same project (Proteus FT, 720p upscaling 200%) runs 793s in 2.6.4 and 1644s in 3.0
PS And please bring back .avs drag and drop support, especially since ffmpeg have versions with internal AviSynth compatibility.
2 Likes
Weāll have to see if devs fix the āknown issuesā to see if the product is really worth it.
Then speed optimization and model optimization should be their first priority.
Like i said in beta forums, i get sometimes a better result with proteus 2 than 3. That is weird.
Hi,
Iāve this issue, with Chrono fast up to 25 > 50 fps around 1/3 more long with 3.0 
(with for me same setting H 264 in MP4, no āenergy save modeā activated)
Power GPU/CPU drop power consumption too 1/4 less.
Win 10
RTX 3070 Ti (undervolt 1860 MHz max 0.9180 gpu voltage)
Ryzen 9 5900x
I donāt have no more time for the moment to try other AI model (upscale and enhanced) Hope there is no issue too 
Is there a reason why there isnāt a model that is trained to both upscale and to interpolate (like Apollo or Chronos or better) - in one go (one pass)? Wouldnāt having a model that was trained to do both in one go probably be faster and better quality than chaining 2 AI models (one for the interpolation and one for the upscale)? eg. couldnāt it be trained by having a video at high res and high fps and creating a version at a quarter of the size and lower fps and having it learn to convert one to the other?
Also in the old Video Enhance AI there was an option to click to view a window to compare 3 different models against the original video. Couldnāt there be an option for that in this new version/app?
In the old version with the compare option (that compared 3 models) Artemis I think a lot of the time looked the best - Iām not sure it is any more with this version and the name seems to have changed to āArtemis - Denoise/Sharpenā. Others have mentioned how the models are different. So have all the models changed even if their names are about the same as before? I donāt seem to get as much detail in an upscaled video as I did with the old version of the app.
Maybe there could be more AI models available that are trained in different ways and some that are optimized for quality and others optimized for speed. Maybe there could be an option for us to train our own AI models.
Hi,
How do I know that the model is using Tensor RT?
I have already uploaded the Log files, 2 photos and a small sample video
PC vs Mac
What is the general consensus regarding the speed of v3.0 versus v2.6.4 on a Mac M1 Mini 16GB machine?
From what I am reading it has slowed down the rendering by as much as 20% on some things. Sorry I havenāt been able to read the 115 posts yet. lol
This concerns me, as I would think speeds would at least remain the same on your new version (at least on the same offerings from 2.6.4).
Is Topaz begging to put more emphasis on the PC version development?
Or is v2.6.4 going to be the fastest you will produce? It appears the majority of your customers are PC users, based on the posts.
How long will you continue to support v2.6.4, as I guess it requires downloads from your servers AI models when it starts to render?
Would this new rebuild version of the app finally support the EXR format? This is critical for professional VFX and CG renders. Also much more efficient than TIFF or PNG.
In the promo page for Video AI v3.0.0 it says āSupport for hardware encoding on all platforms for H264, H265, and AV1 (when available), along with higher-bit output formats such as ProRes 444.ā
But in the Output Settings (and I think the preferences) there isnāt an option for ProRes 444. The output options allow ProRes 422 LT and ProRes 422 HQ but no ProRes 444 even though ProRes 444 was in the advert for this version. Can there please be an option for ProRes 444 (or should it be called ProRes 4444?).
Also can there please be an option for ProRes 4444 with the alpha channel from the source kept? Otherwise it will not be good for upscaling videos that contain an alpha channel where you want the output to be a video.
1 Like
Iām agree with you, for upscale and enhanced
I potentially ended my post wrong and didnāt choose the right example for conclude.
If upscale and other model have change for the new version and do a better result , Iām ok if they take more time.
For a ābasicā x2 FPS the new version does not bring any clear change for basic FPS change (only for slow motion x800 and other).
Only if in made error 3.0 version change the slow motion effect and not the basic FPS change 23.976 > 120 FPS
In my case itās a basic FPS change 25 to 50.
The 3.0 version take around 33% more time for same result (with little more power consumption for same job).
1 Like
Congratulations to Topaz on this release, however I will not download it just yet until the rough edges are smoothed out a bit more. I expect it will be more palatable in a couple of months. Iām a beta tester, not a ārelease testerā. 
5 Likes
On Mac, how do I install 3.0 while keeping 2.6.4. [Edit]: Figured this part out.
Iām really not sure why 3.0 is a Release Candidate while most workflows are slower while using more resources. Other than the UI, is there any thing that 3.0 does better than 2.6.4 that 2.6.4 was doing (other than new models)?.
New Problems. While attempting to de-interlace an SD video the results was a fairly well de-interlaced export imafe. The only real problem with it was the right edge of the imasge has a sawtooth edge. - Does mean the interlaced odd/even interlace lines are different widths, or is ther some kind of mismatch or timing problem when the odd/even lines are reshuffeled during interlace? Secondly, VEAI 3.0 appears to be naming the output image like itās a prefiew file snd ignoring the. It writes into the origin folder.and ignores the file name entered on the Export As screen.
1 Like
I donāt disagree with some of your points, but the negative and unkind demeanor in your posts turns me off, and I suspect Topaz reacts the same way.
1 Like
Same.
They have said clearly, that you can and should keep 2.6.4 around and it will work for anyone that has bought it in the last year. True itās annoying and rash that the installer for 3.0.0 removes it, but it was not hard for me to get 2.6.4 back and running.
Should they have kept it in early access for a few more months? Yes. Does management always judge the state of things correctly? No. I blame the manager that demanded there be an Octobar release of the new version.
1 Like
Great version - However I would wish for better options for setting the Nvidia Codecs in a future version - Like a deeper layer of āexpert settingsā that would be able to stick within the GUI. I can currently only get what I want from command-line utilization.
But it would be nice [and much faster to work with for batch encoding] if the GUI featured more advanced codec NVENC codec settings available where one could specify the codec settings away from the āmediumā and ā-b:vā settings, so it would be possible to specify something like ā-c:vā āhevc_nvencā ā-profile:vā āmain10ā ā-presetā āp7ā ā-pix_fmtā āp010leā ā-cqā ā1ā ā-qminā ā1ā ā-qmaxā ā1ā or even ā-c:vā āhevc_nvencā ā-presetā ālosslessā for use with encoding in the GUI (and likewise for the upcoming AV1 encode support)
Thanks 
3 Likes
Version 3.0.0 didnāt remove 2.6.4 for me (Windows 10). It gave me an option to remove version 2 but I declined it.
3 Likes
I think I have figured it out, but am waiting on confirmation. 3.0.0 always sets the scale of the model to 0. This results in a model for a bigger resolution being selected.
