High Precision AI Smart Mask 2.0 not working

MutantPike wrote on 1/6/2026, 12:47 PM

Hi there,

I purchased Vegas Pro 23 yesterday, hoping for improved performance, especially when using some of the tools.

When I started the programme, I got a prompt informing me about the update for DLM Smart Mask.

I didn't think much of it, but I updated it straight away.

However, once the update was complete, I quickly learned that it had broken AI Smart Masking 2.0.

When choosing "Precision: High, Slow", and starting the "analysis", the pop-up window starts flickering on and off, without end. Alternatively, the pop-up message about the "analysis in progress" just disappears, and the AI Smart Mask 2.0 window looks as if the analysis finished, but on my preview window, I cannot select anything, nothing is being highlighted or whatsoever.

On the other hand, "Precision: Low, Fast" works perfectly fine with somewhat good results. Weirdly enough, I can see the same behaviour in Vegas Pro 22. I guess since both Vegas 23 and Vegas 22 use the same version of AI Smart Masking, the problem applies to both, right?

I've seen that the VEGAS Pro 23 Build 356 patch release, which was withdrawn on December 18th, addresses that: "Fixed problems with Smart Mask when using the high-precision model.".

Thus, I am wondering if there is something I am doing wrong, or if the tool is broken at the moment, and there is going to be a patch/fix/update. (I don't think I am doing anything wrong cause I used it a lot in the past).

Thanks

Comments

RogerS wrote on 1/7/2026, 8:54 AM

Hi Mutant, could you add your system specs to your signature? I'm curious about your GPU model and amount of VRAM.

Could you clarify which builds of VEGAS and Smart Mask you are using?

I just opened 356 with the newest model for smartmask and had it analyze one frame and find an object using high, slow precision. I then had it generate the rest of the frames. I don't see an issue with the window popping in and out here.

This is tested with my RTX 4060 (8GB VRAM) laptop, studio driver 581.57.

MutantPike wrote on 1/7/2026, 9:04 AM

Hi Mutant, could you add your system specs to your signature? I'm curious about your GPU model and amount of VRAM.

Could you clarify which builds of VEGAS and Smart Mask you are using?

I just opened 356 with the newest model for smartmask and had it analyze one frame and find an object using high, slow precision. I then had it generate the rest of the frames. I don't see an issue with the window popping in and out here.

This is tested with my RTX 4060 (8GB VRAM) laptop, studio driver 581.57.

Thanks!

Sure, my specs are:
GPU: RTX 4080 SUPER 16GB.
CPU: Intel(R) Core(TM) i9-14900KF (3.20 GHz)
RAM: 64GB

Additionally, I have just reinstalled DLM Smart Mask, the latest version, to see if this resolves the issue, but unfortunately, it has not. This is what I get as a result:

When I try to select an object, I get a weird smear that is not specific to anything in my preview window (This is using High,Slow precision...

Low, Fast precision works somewhat ok:

RogerS wrote on 1/7/2026, 9:07 AM

You are on build 302 and the current smart mask build released with 356? The older model is here if you want to try it.

Which NVIDIA driver are you using?

MutantPike wrote on 1/7/2026, 9:17 AM

I also just updated my previous response with more details.

I actually just updated the drivers to GRD 591.74 and tried the process again.
As in my previous message, I am getting this weird smear when using High,Slow precision Smart Mask 2.0.

 

You are on build 302 and the current smart mask build released with 356? The older model is here if you want to try it.

Which NVIDIA driver are you using?

 


UPDATE:
Ok, I tried the older model, and it works beautifully using High, Slow precision Smart Mask...

So I guess... Sorry to say that, but... Magix, you broke something, haha.
But I am glad to see that it can be fixed!

UPDATE2:

Although the older model is less problematic with detection,

high,slow precision doesn't seem very consistent in performance. Sometimes it detects something, and some other times it doesn't detect anything, and sometimes it is just stuck analysing.