Video Roadmap Update (Feb 2023)

For CLI processing I remove the “-nostats” so I can see the progress and fps etc.

I also find ffmpeg’s built-in suspend / resume very useful too for CLI processing.



I think this parameter should not be added by default, thanks for the reply.

1 Like

What is going on with the bugs that have been occurring since the v.2.x.x days that are still left unresolved?
There may be calls for shake support, fps changes, etc., but it is fatal that there are things that can cause the picture to break down as a picture.
I am frustrated that models such as Proteus have not been updated at all.

1 Like

If you are implementing face restoration, would this be accessable as a early alpha ? :slight_smile:
Face restoration would be a game changer !


I appreciate all the hard work you’re putting into this new version.
Tired of seeing people complain about the software.
Angry users your Anime and star trek upscales will be ok if it takes time to work out some bugs. lol

1 Like

Are there any plans to effectively remove watermarks from videos? Or is this just a step too far?

This would be a great feature.

  1. Video quality models for upscaling and denoising. While true face restoration for video is currently infeasible, we’re exploring options to improve how natural faces appear when upscaled.

There are some companies who offer this option on the internets, but beyond potential ethical and legal implications, from the purely technical point of view, to be quite honest I was far from impressed what is currently on the market.

HitPaw Watermark Remover
AI Powered Image and Video Watermark Remover

If Topaz were to introduce such a feature, I would expect them to introduce something less niche and more widely applicable, like the “content aware fill” Adobe has, or “Object Remove” feature Blackmagic has, only much better. That would be quite useful.

Ideally something like Mocha Pro replace model, but automated and enhanced with AI.

1 Like

Thank you, I’ll have a look at your link. It does sound complicated to replace what is behind a watermark as the scenes change.

1 Like

Yes. Generally speaking, there were two approaches to remove objects from a video. One was special and the other temporal.

Special one is like when you do it in Photoshop on an image. You clone or heal parts of the image from surrounding area in order to replace object in the image you don’t want. Obviously, this is limiting and time consuming.

Option B, the more advance temporal method, only available in video was to use frames before and after to fetch parts of the image you want to replace. But to make this work you also have to adjust for changes in lighting and color.

Mocha Pro is very good at doing this, and Stabilization model in Topaz tries to do something similar to replace parts of the frame once it has stabilized the image. Because often one has to zoom in or crops in to hide the changes in the size of the frame, you lose parts of the image. But one attempt to compensate is to simply look for areas from before in order to fill in the cropped areas.

This is something Mocha Pro has been able to do, with manual assistance for a long time. Topaz tries to use automation with the help of AI to do it automatically.

Here is a demo of how Mocha does it and how basically Topaz is trying to automate the same idea.

Here is how it works in Mocha when you are trying to remove an object by replacing it with temporal method of borrowing data from frames before and after… and off course adjusting for the lighting and tone changes.

This same idea can be used to extend the background for further VFX or clean-up work.

Sometimes, however, you don’t have a frame in the video where the thing you want to remove is not visible, so you can’t borrow clean frames and use them to erase the object and then you need to manually take one frame, use Photoshop to clone out the parts you don’t want, manually and tell the Mocha to use that as reference for other frames.

But that is a manual process. Ideally, we would have AI which can use the database of images to approximate the background behind the obstructing object and do it automatically, but as you can imagine that is the most complicated version of them all and for Topaz to implement that, might take many more versions I imagine.


Thank you for this info!

1 Like

@guilherme.salles This is absoloutely mind blowing. Everything changes, If TVEAI can get something like this within itself.


This would be fantastic!

1 Like

It’s not feasible because it would be incredibly slow for normal use.


  • bring back the suffix and prefix options for output files.
  • disable momentum on scrollbars. Its difficult to use the scroll wheel.

Any plans to optimize for the Mac M series processors with integrated GPUs?

I definitely support it, since I myself am going to buy an rx 7900 xtx in the future

I hope the speed increases also for the mac m1 version, because it is impossible to work with a big file. Example: any 1-hour low-resolution video taken to 4k takes about 8 hours, unthinkable…


New update: