Whis is Topaz Video 3.x quality so BAD compared to 2.3? šŸ¤¦šŸ»

Many parts look cleaner and smoother (and warmer tones) and many details are just gone, but I assume that is also due to the HVC1 compression, which is not good for a true comparision. Can you export to PNG or any other lossless output in the future? exact frame is 244,59.94, at time 00:00:04.071

2.6.4 Top - Hybrid+3.3.6 Bottom

Look at the monitor frame letters. In Hybrid+3.3.6 is just a blur. Same with the notebook. All the detail is gone: no lines in the page and the spine is also a blurry mess.

2 Likes

Any news on this problem?

Can you double-check the FFmpeg command from both exports to ensure that they are the same?

I think I already provided plenty of tests, analysis and sources for you to make your own tests.
I donā€™t think this has to do with the encoding part.

@ida.topazlabs
Iā€™m pretty sure heā€™s reporting on this same issue that I and others have reported several times.
Topaz Video AI v3.0.0
Topaz Video AI v3.0.0
Topaz Video AI v3.0.2

I remember making a more detailed report, but couldnā€™t find it.

you mean this?

No. That oneā€™s specific to Theia.
It had pictures of the same show, but was pointing out that a different model size gets used in TVAI than VEAI with the same settings. And that 1X, 2X and 4X all produce very different looking results. My conclusion was that TVAI picks the 4X model where VEAI picks the 2X modelā€¦ At least with Artemis High Quality upscaling a DVD source to FHD.

Yes, that is what our thought process is with asking the OP for the comparison of the FFmpeg commands.

Thank you!

So in other words you saying, TVAI out of the box is optimized for x4 upscaling while VEAI to x2 upscaling?

No. Optimized sounds like it was intentional. I definitely think the difference in scale selection between the two is in error.

There might even be models that are now inaccessible in TVAI because it always chooses the 4X model, or something like that.

OK, Iā€™ll put it differentlyā€¦ not optimizedā€¦ but you suggesting that TVAI does a better job for x4 upscale and VEAI does a better job for x2 upscale (rather intentional or a fluke)?
in other words if I need to upscale by x2 I would probably have a better chance/results in VEAI ?

I suppose I may have implied that, in a very roundabout way.
When using VEAI, I thought that Artemis HQ was perfect for upscaling a certain DVD series to FHD.
It was using the 2X version of the model then using a simple upscale filter to do any more needed upscaling.
When I tried the same DVDs and settings in TVAI, it gave the same results as VEAI for when I told it to do a 4X upscale then downscaled it to FHD. I then figured out how to modify the command to get it to pick the same 2X model as VEAI. In that way, I get the speed improvements of TVAI and the results I wanted.
So for, VEAI doing a better job for x2 upscale? Only in very rare circumstances, and if found and reported to Topaz, all can be corrected and done in TVAI, eventually.

can you give me a full command example? I want to test it to see if i spot any difference between the modified command and x2 upscale through GUIā€¦
I want to try that command trick if you donā€™t mind sharingā€¦

Hereā€™s the post I meant to make back when TVAI first came out. (But with the latest version since itā€™s still an issue.)
With a DVD source in VEAI, I select FHD output and Artemis HQ as the model.
Same input, same output in TVAI. This is how they look:
image
Left VEAI. Right TVAI.
Now I take that same video in VEAI and up it to 4K then down it back to FHD:
image
Now I take the the command generated by TVAI for the FHD version,
"G:\Program Files\Topaz Labs LLC\Topaz Video AI\ffmpeg.exe" -hide_banner -i "E:/Dump/HOGANS_HEROES_D1/A2_t01 ff.mkv" -sws_flags spline+accurate_rnd+full_chroma_int -color_trc 2 -colorspace 2 -color_primaries 2 -filter_complex scale=w=640:h=480,setsar=1,tvai_up=model=ahq-12:scale=0:w=1920:h=1080:device=0:vram=1:instances=1,scale=w=1920:h=1080:flags=lanczos:threads=0:force_original_aspect_ratio=decrease,pad=1920:1080:-1:-1:color=black -c:v png -pix_fmt rgb24 -start_number 0 "E:\Dump/A2_t01 ff_ahq12/%06d.png"
and I change the scale to 2.25. Itā€™s the first scale after tvai_up. Above itā€™s 0. Below itā€™s 2.25.
"G:\Program Files\Topaz Labs LLC\Topaz Video AI\ffmpeg.exe" -hide_banner -i "E:/Dump/HOGANS_HEROES_D1/A2_t01 ff.mkv" -sws_flags spline+accurate_rnd+full_chroma_int -color_trc 2 -colorspace 2 -color_primaries 2 -filter_complex scale=w=640:h=480,setsar=1,tvai_up=model=ahq-12:scale=2.25:w=1920:h=1080:device=0:vram=1:instances=1,scale=w=1920:h=1080:flags=lanczos:threads=0:force_original_aspect_ratio=decrease,pad=1920:1080:-1:-1:color=black -c:v png -pix_fmt rgb24 -start_number 0 "E:\Dump/A2_t01 ff_ahq12Fix/%06d.png"
Now hereā€™s the result:
image

I have tried to change the scale in the command on other models. Some work. Some do not.
Hereā€™s the files. Enjoy.
theFiles.zip (14.9 MB)

3 Likes

I am confused a bit, maybe i misunderstood you.

then figured out how to modify the command to get it to pick the same 2X model as VEAI. In that way

If I load a video in TVAI (720x540) all I need to do is change in the command the value scale=2.25 and it would supposedly use 2x model as VEAI?

Depends on the model. I know for sure it works on Artemis HQ, but it doesnā€™t do anything on Artemis Strong Dehalo. I have not tried it on many other models.
2.25 is just what VEAI said it was increasing the resolution by. You can put other values in there.

The only way to know for sure is trial and error using both TVAI and VEAI.

I tired 2 clips at 720x540 resolution. I decided 1st to use the UI defaults to see if there is any difference before I try tweaking the command as you suggested.
I loaded them both in TVAI and in VEAI. I selected upscale x2 (1440x1080) and Artemis HQ (to match your bed test) and I used Video-Compare tool to view the differences. there was no difference in the video quality between the two. I tried the same with Artemis MQ and Proteus and they were identical.
So i decided there was no point to run those with a tweaked command prompt.

Maybe it happens only on very certain videos.

I donā€™t get most of what is being stated in these last posts, but one of the keys of my tests is that the sources are interlaced, so the thing is that it seems VEAI does a better quality deinterlacing process than TVAI.

@james-4065 @bombayetas @whophd @ForSerious @alanhusband-677420 @Akila Thank you all for your feedback here. This is something that will take a bit longer to resolve on our end.

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.