Topaz Video AI v3.2.1

Despite all the talk about versioning … standard practices are: (Source Thales)

Some of the most common methods of software version numbering are:

  • Semantic numbering – Three-digit numbering technique based on Major.Minor.Patch
  • Date-of-release – The software version number is the date of the release. For example, 20.06 (June 2020)
  • Unary numbering
  • Alphanumeric codes
  • Sequential numbering

Topaz clearly use semantic versioning which is generally considered best practice.

2 Likes

v3.2.1 crash while almost finishing after more than 15 hours processing: “Unknown Error”

I initiated 4 videos with running 3 in parallel, and the 4th waiting. For more than 15 hours everthing run ok, temp files continued to get updated. Than those 3 stopped with “Unknown Error”, all 3 temp files disappeared, and the 4th started processing, as wanted with its temp file continously updated.

A way to intelligently detect high motion scenes (maybe allow us to select minimum “scene” length in seconds) and then use frame interpolation only in those areas, and just use cheaper frame doubling in other areas would be cool.

Hello, is there anyone got this ? I see that after I preview AI is have a problem with subtitle, it not clear, it have wrong, can anyone help me this ? I use Video Enhanced version 2.0.0, anyway I also try this new version 3.2.1 but it same, anyone have solution for this ? Thanks!

P/s: I try to test with video just 144p on Youtube - how can I do upscale for seeing everything be sharp and clear ?

I write again!!! Sorry, but my processor is busy with other programs. I set in the TVAI settings, use a video card. But he uses it by 70-80%, and the processor loads completely. I now have a processing speed of 2 videos
in 4K - 0.3-0.8fps. That is, TVAI processes 40 seconds of video 2.5 hours and 8.5 hours.

Two things: 1) just enabling video cards doesn’t mean that Topaz is only using the video card. There is still plenty of work that it requires to be done via CPU. If this is truly destroying your workflow you can try changing the priority at your system level but be warned that this will likely just result is even longer encode times.

Speaking of longer times, what are your PC specs? My guess is that your main constraint is the power of both CPU/GPU more than anything else

if you upscale a 720p wide or smaller video the subtitles get messed up. as a temporary workaround scale up the video to 1280p width with a video converter of your choice before processing with video ai!

I have an i712700, NVIDIA 3060. Previous firmware had GPU 95-99, CPU 30-40. Now it’s 80/80

I am still unable to download all models for my RTX 2080 SUPER.

I could retreive most of the modells from my old installation from prior the download problems.

However the Apollo I didn’t have and I can’t download it.

My system specs:

Topaz Video AI  v3.2.1
System Information
OS: Windows v11.2009
CPU: AMD Ryzen Threadripper 3970X 32-Core Processor   127.9 GB
GPU: NVIDIA GeForce RTX 2080 SUPER  7.8135 GB

Do I need to provide more information?
Thank You.

Proteus auto looks better now.

1 Like

I’m processing 640 to 1280 with Proteus Auto currently. I had a RX 6600 but it was just too slow to use effectively.

Topaz Video AI  v3.2.1
System Information
OS: Windows v11.2009
CPU: AMD Ryzen 9 3900X 12-Core Processor              63.952 GB
GPU: NVIDIA GeForce RTX 3060 Ti  7.8496 GB
Processing Settings: device: 0 vram: 0.95 instances: 1
Input Resolution: 640x480
Benchmark Results
Artemis		1X: 64.36 fps 	2X: 44.84 fps 	4X: 14.14 fps 	
Proteus		1X: 64.28 fps 	2X: 44.01 fps 	4X: 14.25 fps 	
Gaia		1X: 23.30 fps 	2X: 16.52 fps 	4X: 11.33 fps 	
4X Slowmo		Apollo: 67.15 fps 	Chronos: 51.31 fps 	Chronos Fast: 67.37 fps 	

Topaz Video AI  v3.2.1
System Information
OS: Windows v11.2009
CPU: AMD Ryzen 7 5700X 8-Core Processor               63.893 GB
GPU: NVIDIA GeForce RTX 3060  11.845 GB
Processing Settings: device: 0 vram: 1 instances: 1
Input Resolution: 640x480
Benchmark Results
Artemis		1X: 57.37 fps 	2X: 43.17 fps 	4X: 14.76 fps 	
Proteus		1X: 57.13 fps 	2X: 41.59 fps 	4X: 14.09 fps 	
Gaia		1X: 18.96 fps 	2X: 13.30 fps 	4X: 9.15 fps 	
4X Slowmo		Apollo: 53.70 fps 	Chronos: 42.50 fps 	Chronos Fast: 55.84 fps 	

Topaz Video AI  v3.2.1
System Information
OS: Windows v11.2009
CPU: AMD Ryzen 5 5600X 6-Core Processor               31.906 GB
GPU: NVIDIA GeForce RTX 3060  11.845 GB
Processing Settings: device: 0 vram: 1 instances: 1
Input Resolution: 640x480
Benchmark Results
Artemis		1X: 57.14 fps 	2X: 43.26 fps 	4X: 14.43 fps 	
Proteus		1X: 57.06 fps 	2X: 39.86 fps 	4X: 14.17 fps 	
Gaia		1X: 19.24 fps 	2X: 13.46 fps 	4X: 9.36 fps 	
4X Slowmo		Apollo: 58.85 fps 	Chronos: 42.89 fps 	Chronos Fast: 58.19 fps 	

2 Likes

I can’t generate preview with the Theia model.

This also happens to the split view slider, if its moved during playback when it reloops the video it resets the slider to the center, really annoying to have to keep moving it over and over.

Semantic versioning:

A normal version number MUST take the form X.Y.Z where X, Y, and Z are non-negative integers, and MUST NOT contain leading zeroes. X is the major version, Y is the minor version, and Z is the patch version. Each element MUST increase numerically. For instance: 1.9.0 → 1.10.0 → 1.11.0.

If the text is not readable on your source video, TVAI can’t know what the text is supposed to say, it just sharpens what’s there.

Going way back to the initial release, fonts have always been a problem. TV/Movie credits are a mess and always have been. It was suggested at one point to have a model dedicated to text/fonts, but it never went anywhere. Useful suggestions tend to get trash binned anyway.

2 Likes

Same issue since V3.2.0.
System specs: win11 22h2 + rtx3090

Is it best to upscale via another program first (eg Handbrake) then enhancing via Topaz? Someone did allude to this earlier in the thread and I wondered why that is?

I wouldn’t touch Handbrake with a barge pole for upscaling.

2 Likes