Topaz Video AI Early Access v3.0.0-2

Hello Everyone,

We have another update for Topaz Video AI 3.0 with lots of fixes and improvements. Please provide your feedback.

Windows | Mac
Headers, Dylibs, & Libs for custom FFmpeg builds

Changes from v3.0.0-1

  • A new Apollo model is added for 4 to 8x slow-mo to better handle non-linear and large motions
  • Close all added for export
  • Improved image sequence preview/export playback
  • Output encoder can be modified by the user externally
  • Mouse draggable trim
  • Export can be previewed after enough frames are processed, if allow recovery is ON in preferences
  • Faster loading/updating videos
  • Fixes issues with some filter chaining
  • Add warning when video failed to load
  • Filename contains filter information
  • Metatag videoai added to videos with full filter details
  • Fixed looping while processing
  • Misc. fixes and improvements

Known Issues

  • Apollo may fail on some GPUs if the device setting is Auto
  • Gaia on Mac may cause failure
  • Preview mismatch for certain VOB files
  • Some videos can still have some seeking issues in Preview/Export
  • Color space / HDR issues
  • Keyboard navigation missing
  • Intel codec not working on some machines
  • Some quality mismatch between 2.6.4 and 3.0
  • Multiple selection for source videos is missing
  • Output bitrate is not reset to Preferences bitrate when importing a new file
  • Different view mode is still missing
  • Failure can happen when resizing to a big resolution like 4k or 8k without a warning if machine runs out of memory
  • Speed can be slower than 2.6.4 with some models on some machines. We’ll be working on speed optimization in the next weeks
  • Proteus with Auto + Stabilization may cause artifacts
  • Preview with image sequence doesn’t work or plays at 25fps

Problems & Known Solutions

  • If the output doesn’t open in other applications, disable Allow Recovery before processing in Preferences > Output (note: Allow Recovery is required for previewing exported video in the app while processing)
  • Some Mac users who participated in the Alpha or Beta may experience failures with all models. If this happens make sure the data directory is pointing at the correct version of Video AI
  • H.264 encoders have a maximum resolution of 4096x4096, use a different encoder for 4k or 8k videos
  • Changing data folder path can cause issues. To resolve: manually create a folder and name it models in the new location and copy the files from old to new location

Proteus Auto Caveats
The adjustments will not be visible and the sliders will show values of “0” when in auto mode. However, each frame is being analyzed for optimal settings while processing. If you keep Auto enabled and modify the sliders, the settings you choose will be value-adjustment percentages. For example, if Auto estimates the Sharpen slider with a value of 50, and you manually select -50, then the actual value applied will be 25.

CLI Usage
CLI mode has returned for Topaz Video AI 3. Detailed usage instructions are available here, on our support website .

Frequently Asked Questions can be found here .

Please upload videos that are not working or having any type of conflicts to this DropBox Folder .


(possible) bug : at first installation, after start up, setting in preference are not keept immediatly.

1/ set up : 12mb/s - H264 , mp4. no customise output set.
Export → Mov 12m/s.
2 / set up customise output to Mp4 12mb/s. export is mp4 12mbs
3 / turn customise output off : Mov 12mb/s is back

4/ leave Veai and launch again
5/ go in preference : Mov and 12 mbs are back there.
6/ set up in preference again Mp4 and 12 mbs
7/ Export work ok in Mp4 12mbs :slight_smile:

so preference must be set twice and opening the software twice (after installation) ?

  • Preview start with a static image preferred
  • Individual single frame selecting and preview preferred.
  • This would speed up a lot generating the preview
  • Precise cutting of frame number start and end preferred.
  • Any preview resets the zoom level.

Great stuff so far. I’m very happy to see the ffmpeg integration and piping, video cropping, and especially the standalone frame rate adjust option! Running previews as a separate process – amazing idea, A/B testing just got way easier. All of these things will cut down on processing time immensely.

Some feedback:


  • If a preview is left to loop for a little while, the program seems to crash. Has happened 3 times (After a few minutes left looping @ 5 seconds)
  • The folder preferences reset when relaunching the program.
  • The window does not attach or go full screen when dragged by the bar to any edge of the monitor on Windows 11
  • Sometimes when the preview catches up to the last processed frame, it will briefly skip to the first frame then back. Sometimes this results in the previews getting a bit out of sync until you skip somewhere else in the preview.
  • The current frame # (or any frame #s) are not displaying anywhere that I can see.


  • Agree with Imo on being able to select start:end frame for trimming video, like in VEAI 2.x. Also a more obvious start:end marker on the playback bar.
  • A slider to adjust the duplicate frame dropping power would be great (if its mpdecimate, idk). On compressed videos, some duplicates fail to get dropped due to compression artifacts. This results in stutters in the output, especially if using slomo or increasing frame rate.
  • I think the preview sizing options (fit, 1:1, 1:2, etc.) should be displayed above the playback bar by default. I was going into the menu to change that for a while until I finally clicked on the magnifying glass and realized they were hidden there.
  • There needs to be a way to stop a preview from processing once it’s started without deleting it entirely.

Wish List:

  • The ability to preview videos at a high bitrate (ie 4k60, 1080/120, etc.) if the user’s monitor supports it (maybe separate window?)
  • GFPGAN integration, if possible. Faces don’t always do so well in VEAI, or can require a lot of trial and error with low-quality video. I suppose you can do it in ffmpeg, but it would be nice to just check a box and have it in the preview.

Thanks for all the hard work! I’m very excited to see where VAI goes.

Edit: nevermind on the GFPGAN integration. I ended up writing a program to do it, and it is not temporally stable at all lol. The results range from decent to terrifying. Guess I’ll just have to wait another few years. Would be nice to have a model trained on resolving motion blur too.


Just tried to run and ran into a failure immediately:

expacc 127 is out-of-range
error decoding the audio block
ignoring pic cod ext after 0
Input #0, matroska,webm, from ‘THE_LIBRARIAN/The Librarian Quest for the Spear.mkv’:ignoring pic cod ext after 0

I was looking to clean up the video with Artemis - denoise/sharpen Normal Quality and no scaling (original size)

3.0.0.-2 works fine with only one File imported. That can be processed 3 times in parallel with 3 different settings. I’m impressed by the crop function!

However mutliple files cannot be imported. How to achive 3 differnt files to get processed in parallel?

Hi, i assume you’re talking about launching different files at the same times to get them processed ?
the feature to select multiple files to be processed, from what it has been told to me, will be available on on friday beta release.

meanwhile, if you want to have multiple videos processed,

  • click on one video, do your settings, and check Export, click on close, select next video do your setting (unless it has been done before) and click again on export.
  • etc …

all the videos will be but in the export list.
if in preference, the number of parallel process is set to one, it will do one video at a time in the order of the list (from bottom to top), first video is processed, and others are put on hold.
if there are more process (2 or 3) in preference, it will process them in parallel at the same time.


Thank you

1 Like

3.0.0-2 VEAI crashed after 68% work done, but ffmpeg.exe is still running with 40% CPU an 30% GPU running

1 Like

can’t reproduce it, but on some videos, trim turn on itself alone. after importing a bunch of video, and going from one to one to set it up with proteus auto and make different preview, one (already 2) has the trim in/out turned on in orange, but i didn’t activated it, and with the end set up a bit before, (the left locator is set at 00:00:00).

like that it keep now the zoom and window position of the preview, we can go from a preview to an another by almost keeping the same original position (a lock button for the position and zoom, would be best, so we can watch each preview with the same visual setting).

noticed an undo icon in the trim in/out box.

pure personnal suggestion maybe not shared by others : would prefer to set the in/out by directly clicking the timineline with mouse instead of doing a click and hold of the left and right bracket.

  • having the left bracket set up automatically to where the cursor is when we take a look at the video / preview before turning on this feature.

If I try to use the “Deinterlace”-Filter combined with the “Enhance”-Filter I always get en Error “Download Failed”

Here is a Log-File…
Could you please help me?

2022-09-01-08-39-58-Main.tzlog (492.9 KB)

or rather it occurs in almost every constellation of Filters! :rage:

does anyone know how to change the encoding speed in the encoder.json file from medium to ultrafast? I tried but it doesn’t work it refuses ultrafast h264, h265, prores etc please help me

Don’t know if Nvenc presets are part of the Topaz compiled version of FFmpeg but the Nvenc encoder does not use the same naming structure or implementation as h264/h265. The presets even change depending on generation of video card in use. All the cards appear to default to a medium setting for whatever that’s worth. You can do a search on google for ‘nvenc preset names’ and also check out the Nvidia NVENC Preset Migration Guide.


Thanks for your response @MikeF MikeF
I hope that will change later, it would be nice if we could really change them. :slight_smile:

1 Like

Gaia HQ download fails:

Failed to configure output pad on Parsed_veai_up_0
Error reinitializing filters!
Failed to inject frame into filter network: Invalid argument
Error while processing the decoded data for stream #0:1
Conversion failed!

Other models download and preview fine, including Gaia CG, Artemis and Proteus
2022-09-01-11-17-0-Main.tzlog (148.9 KB)
2022-09-01-11-09-57-Main.tzlog (45.4 KB)
2022-09-01-11-05-54-Main.tzlog (7.6 KB)

UPDATE: It seems only the 2x Gaia HQ model download fails. 1x and 4x seem to be fine.

are you on mac ? this is written in changelog :, don’t know if it’s related :

  • Gaia on Mac may cause failure

Ok, yes!

One more question, unrelated: Mouse draggable trim, does that mean I am able to crop the video with the mouse?.. because it doesn’t work for me.