Topaz Video AI v3.1 – faster exports, faster playback, and new Motion Deblur model

[eric (Eric Yang)]

Topaz Video AI v3.1 offers significant performance improvements and introduces a new AI model Themis that sharpens your video. Get it here:

Topaz Video AI v3.1.0

Hello Everyone! We have a new release of Topaz Video AI with significant performance improvements and a new AI model. v3.1.0 Released January 17th, 2023. Download: Windows , Mac . Changes from v3.0.12: Faster processing for all models New AI model for Motion Deblur (Themis) Faster preview playback Chronos, Chronos Fast, and Full Frame Stabilization uses TensorRT models on NVIDIA machines Noise and Grain settings are same as v2.6 All filters and environment variables have been renamed to t…

150-400% faster previews and exports

You can expect significantly faster exports and previews from Topaz Video AI v3.1. The exact speedup depends on your hardware, video, and AI models used, but internal testing has recorded a consistent +50-100% performance improvement on most configurations.

Internally, we’ve parallelized the processing pipeline to improve hardware efficiency. We have not reduced or optimized the AI models themselves, so model quality should remain identical to earlier versions.

Mac M1

Nvidia

Chronos, Chronos Fast, and Full-frame Stabilization also now utilize TensorRT, which further increases processing speeds up to 3x (and up to 6x compared to the last 3.0.x release) on NVIDIA machines.

Get sharper videos with the new Themis AI model

Reduce motion blur caused by panning, rotation, or zoom with the new Themis AI model:

Topaz Video AI v3.1 - Motion Deblur Example

Because it can use information from multiple frames, the new Themis AI model performs much better than its equivalent in our image apps. It also pairs well with the Stabilization model to reduce camera shake and jerkiness:

TVAI Themis and Stabilization Example

Themis helps you increase the perceived sharpness and detail of your videos by removing blur instead of just increasing edge contrast. Give it a try and let us know what you think!

Faster in-app playback

The in-app video player will now play videos at up to 3x faster, as we’ve improved the efficiency of frame rendering. While playback may still not be real-time for large videos, it has significantly improved over v3.0.x and will improve more in the future.

Other quality of life improvements

The 6 releases over the past month have also improved the general quality of life of the app:

• Improved JPEG output quality
• Prevent OS from sleeping when app is processing
• Fix many issues related with image sequences imports
• More reliable login method
• Improve stability and fix various crashes

The road from here

We’re releasing improvements quite frequently, and we’re just getting started! Please give us your thoughts on the Topaz Video AI roadmap to help us guide our development:

Video Roadmap Update (Jan 2023)

Since the last roadmap update, we’ve substantially improved export performance, sped up in-app playback, and added a new Motion Blur model: With this performance update, we’ve revised our core focus areas: Full speed ahead on video quality models for denoising/upscaling. Afterwards, explore options for Face Recovery (1). Continue building a smoother in-app previewing experience, particularly scrub performance and the ease of comparing before/after results. Now that we’ve shipped the major …

---

Thanks to Brian Matiash for examples and benchmarks used in this article.

[david.123 (New Profile Picture!)]

Congrats to the team at Topaz for this release!

Video AI is getting better and better!

(Also, first like again).

[Moebius]

I was just complaining about the speed in the previous version’s forum page. I guess I spoke too soon! Can’t wait to try this out. The speed upgrades sound great!

[michael-3397]

I’m not seeing Themis in the AI Model dropdown, only Theia. On a MacBook Air M2. ??

[michael-3397]

Never mind, found it!

[StamJ (JSS)]

Still not working correctly for jitter and stabilization.

[richardcrawford0]

Updated to 3.1.0 from the prior 3.x version and it tells me I must renew my license. This surely must be an error since this is just a minor update, right?

[20rushtonj (Alaska)]

From my understanding minor updates are things like 3.0.X.
Updates from 3.0.X to say 3.1.X are considered major updates. But I may be wrong about this. So wait for an official confirmation from an employee at Topaz Labs.

[lemans99]

When i try to browse .png video frame snapshot i receive this error:

“Error loading 1 file”

The system is Win10_64

[icc.kl.chan (Banjamin)]

I wonder if anyone could reproduce a bug with AVFS files
So I know some people here pre-filter before inputting their video in TVAI through avisynth/vapoursynth
and with this version I cant seem to open AVFS’s .avi files in TVAI
It will instantly crash so I’m wondering if anyone else has this issue

[dtischler]

I got the same problem i cant use 3.1 because my license expired. So they might see this as an major update. What is really bad is that i now have 3.1 installed which i cant use and i cant download 3.0.12 from the website. Only the latest and 2.6.4. Is there a way to download the older Versions of 3.x?

[41689592]

Many models had problems

Model loaded 0s - > unknow error

ghq gcq was a disaster area

gpu 1060 6g
cpu 1600x
dirve 528.2

ffmpeg “-hide_banner” “-nostdin” “-y” “-nostats” “-t” “15” “-i” “F:/sd83.mp4” “-flush_packets” “1” “-sws_flags” “spline+accurate_rnd+full_chroma_int” “-color_trc” “2” “-colorspace” “2” “-color_primaries” “2” “-filter_complex” “tvai_up=model=ghq-5:scale=0:w=3840:h=2160:device=0:vram=1:instances=1,scale=w=3840:h=2160:flags=lanczos:threads=0:force_original_aspect_ratio=decrease,pad=3840:2160:-1:-1:color=black,split=2[out1][out2]” “-map” “[out1]” “-c:v” “tiff” “C:/Users/li/AppData/Local/Temp/778614161/temp/_tvai_12841924/%06d.tiff” “-map” “[out2]” “-c:v” “hevc_nvenc” “-profile:v” “main” “-preset” “medium” “-pix_fmt” “yuv420p” “-b_ref_mode” “disabled” “-b:v” “0” “-map_metadata” “0” “-movflags” "use_metadata_tags+write_colr " “-map_metadata:s:v” “0:s:v” “-map_metadata:s:a” “0:s:a” “-map” “0:a” “-c:a” “copy” “-metadata” “videoai=Enhanced using ghq-5. Changed resolution to 3840x2160” “C:/Users/li/AppData/Local/Temp/778614161/previews/sd83_ghq5.mkv”

[DmitryV (webhlam)]

It’s good that the performance has improved and a new model has been added!
But on long videos stabilization still does not work (the model analyzes the whole recording for movements, and then the image - squares)! It works with cut pieces for 3-5 minutes, maybe 10-15 without frequent frame changes and movements, but on long videos it is. And cutting video before editing is not the most convenient option. Maybe somehow you can solve the problem?
And add turning off the PC after completing tasks! Thanks!

[marroh]

Hello,

would be really nice to have another (beta) update, but unfortunately I can’t even test the latest update!
When I am starting the software I only get the following error:

“Licence expired on Sat Dec 10 2022”

Luckily v2.6.4 is not affected by the license bug. This leaves me with at least one working and stable version of the software.

[Dilson]

Congrats to the team on this release. I am seeing great speed improvements. Keep up the good work.

[vktopmanz (topmanz)]

Great to read that v3.1 improves the speed of exports. Much nicer not to have to run multiple instances to get better performance out of this software.

I have question on behalf of those who compile their own ffmpeg to get more encoder options, when will the redistributables be updated to support Topaz v3.1?

[ian.craig]

I have two 8 core machines - 1 AMD, 1 Intel. I have not got any version of this program to work on either machine after version 2.6 This one generated this message on the AMD machine

Last FFmpeg messages:

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 2 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 1 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 2 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 2 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 1 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)

Last message repeated 2 times

nal_unit_type: 1(Coded slice of a non-IDR picture), nal_ref_idc: 2

[Steverd]

Confirm - NICE speed increase (GTX 3080)

BUT still No improvement for CG
Any advance anyone? I still have to use v3.6.4 for CG

[luciovids]

Trying out the stabilization + deblur option and it is definitely much faster in this model. Hoping the stabilization is better than previous versions which were a hit and miss with my attempts. Thanks guys for working on processing speeds.

[christian.young-6942 (BobOki)]

Mine made me log back in, but I had no problem with my license.