Improved Export Options (July 2023)

Improved export flow, audio options, and image sequence options

Selecting export settings for your output video should now be more straightforward, with several new features and improvements:

  • You can select desired export encoders in a more logical way instead of choosing from a single large list.
  • Image Sequences are now separate from Video exports, support different bit depths, and allow starting filenames from a specific frame number.
  • You will see better messaging when certain combinations of video and audio settings aren’t supported.
  • You can now use the ProRes Proxy profile for MOV exports, and the MKV container can now be used for VP9.

Exporting also now supports more audio options:

  • You can now specify audio codec (AAC, AC3, PCM) and bit rate for transcoding, now renamed “Convert”. Audio codecs will only be visible for supported containers.
  • “Auto” mode has been removed. Previously, Auto mode would choose to Copy if possible or otherwise Transcode to 128k AAC, which sometimes led to reduced quality.

We will continue improving the import/export flow and overall experience for the app in the next few months, so please let us know what you’d like to see next.

Other improvements

Since June’s update:

  • Improved video player playback speed
  • Fixed missing frames when exporting with ProRes
  • Fixed occasional dropped frame when exporting full video
  • Fixed Iris artifacts on Apple M2
  • Added DPX and 3GP import support
  • Added “Live Preview” toggle to Export Settings, reword messaging in Preferences, and add tooltips
  • Added ability to set custom start frame filenames when exporting image sequences
  • Fixed image sequence import/export issues related with mismatching filename and color space
  • Fixed “Back to Input” behavior, which now returns you to the input video that matches the currently selected output
  • Fixed invalid output when exporting PNG/DPX to VP9
  • Fixed “videoai” metadata tag missing on exported videos
  • Various other fixes and improvements

For the complete list of changes, please see the release threads for v3.3.5, v3.3.4, v3.3.3, v3.3.2, and v3.3.1.


We plan to deliver several new major features this year:

  • New 1x denoising model that improves low light performance for high quality input videos
  • Integration with popular NLEs
  • Allow offloading processing to a cloud backend
  • Improve the Iris model for low and medium quality input videos
  • Allow pausing and resuming processing
  • Allow applying a second Enhancement pass without exporting and re-importing
  • Improve the experience for comparing between different previews

If you want to test these features out before they’re generally available, please apply for the video beta program. Thanks for using Topaz Video AI, and we look forward to hearing your feedback!


Thanks a lot, Dev Team, for these improvements!
@eric @gregory.maddra @nipun.nath @ida.topazlabs @tony.topazlabs

Now all that is missing is the ability to select the desired audio track(s) to be converted when the source video (in the various formats Avi, mp4, mkv etc.) contains more than one different audio track!

I have videos in the archive that have 2 audio tracks, the first in Japanese, and the second in my language… and if I run the conversion with the program, it automatically converts the first track (in order of importance in the video file), which is Japanese precisely, while I would like to be able to select the second track (the one in my source language).

Ps. it would also be nice to have an improved handling of the audio downmix when there are multichannel audio tracks higher than the classic 2.0 “Stereo” (e.g. 5.1 tracks also recorded by various camcorders of the past), as is done for example in the Handbrake program!


Thanks for your feedback @JakSpoon, we appreciate your active contributions to our community :slight_smile:

We’re working on options for audio track selection and we hope to have a more detailed timeline for the release of that feature soon.

As for 5.1 downmixing, we’re currently using ffmpeg’s built-in algorithm for 5.1 → Stereo downmixes, which is based on the ATSC Standard for digital audio compression.

This method takes the front-left and front-back tracks and mixes it into the left channel along with the center channel, and does the same with the right-side elements. This does eliminate the LFE channel from the mix, but in general it is recommended to keep the subwoofer element out of stereo downmixes because it can create an overall bass-centric mix that sounds muddy on stereo equipment.

If you have any suggestions on improving the quality of the stereo transcode option, we’d be happy to take a look.

Here’s some more infomation on ffmpeg’s -ac filter.


Since it has not been raised in the list of Next, I will repost the bugs that have been reported in the past.

Have we found the cause for the alteration of color and the appearance of irregular black spots?
If the cause has been found, shouldn’t a bug-fix version, such as Proteus 3.1, be released first before a new model (1x denoising model) is created?
If the cause has not been found, there is a concern that similar bugs will occur even if a new model is released, which will only result in unnecessary costs.

1 Like

Fixed bitrate is not a good option, you need to provide a way to use parameters like crf. Or in a more free way, provide an input box for advanced users to supplement command parameters.


CRF encoding can be possible only with the SOFTWARE encoder, like x264 and x265… but unfortunately they can’t be added into program for Licenses issues!

I hope in the future… i hope! :melting_face:


Thanks @tony.topazlabs for the reply!

I am not very technical in the audio area, but for me, copy as much as possible from the audio management and audio encoding from the program “Handbrake”… the management of multichannel audio tracks is done to perfection by the program (also totally based on FFMpeg)

@q91632 @JakSpoon It’s true that we cannot directly add libx264/x265’s CRF encoding modes, but we are exploring options for adding NVENC’s p1-p7 CQP bitrate control options, which would be closer in performance to CRF mode.

Intel offers a similar adaptive bitrate control in their QuickSync encoders, so we will also be looking into options for moving away from fixed bitrates on that platform.


GREAT news!! :heart_eyes: @tony.topazlabs
I hope then that this new Hardware coding will come as soon as possible…

Ps. Is this new mode the same one that is used by OBS Studio for the excellent quality of its recordings, taking advantage of the capabilities of the latest Nvidia RTX cards?

Here’s the details:

PPS. Other documentation:


The cq parameter is one of the things I said, whether it’s an option or an input box, it’s better than nothing.

Yay especially to: NLE integration, second Enhancement in single run, comparing different previews!


Greatly appreciate the weekly updates.

@Eric Can Video AI When it Exports Frame Files (TIFF,JPEG,DPX,ect) add extra Meta Data into the Frames that PhotoAI can process ?