Topaz Video AI Beta 5.2.0.0.b

Video AI Video AI Beta
1

Hello everyone,

Today we’re excited to release Video AI 5.2 Beta!

5.2.0.2.b

Changelog from 5.2.0.0.b

  • Fixed missing swap views button.
  • Restored colorspace selection.
  • Ensures colorspace correctness for input videos with complete colorspace tags.

5.2.0.0.b

This release includes the new Rhea Enhancement model, Pro seat management, Frame Interpolation for After Effects, Alpha layer support, and many UI and backend fixes.

  1. Rhea Enhancement Model

Rhea represents a combination of Proteus and Iris. The model is intended to be more accurate in preserving fine details than Iris, while also handling text in a less destructive way. Rhea internally processes inputs at 4x scale and then downscales to your selected output resolution.

We’re very excited to have testers working with Rhea. Thanks as always for your feedback and testing.

image
image1920×747 58.6 KB

image (9)
image (9)1920×550 27.9 KB

  1. Frame Interpolation for After Effects

Today we are also launching Frame Interpolation in Adobe After Effects. This plugin addition includes access to all Frame Interpolation models for up to 16x slow motion conversion.

For more information on using the After Effects Frame Interpolation plugin, see this post in the Video AI Plug-ins forum section.

  1. Alpha (transparency) layer support

Video AI can now copy alpha layers from inputs and merge them back with AI model output videos. Alpha layer copying is supported in the following codecs:

  • QuickTime Animation
  • TIFF
  • PNG

image

Pro seat management

Video AI Pro will be launching next week on July 10th. This new license tier offers seat management, commercial usage rights, and multi-GPU support for teams working with Video AI.

As part of this transition in licensing, we will be discontinuing the previous implementation of multi-GPU rendering for the standard license. Our team found that less than 2% of users made use of the multi-GPU setting, and further development will be focused on multi-GPU optimization for Pro teams.

If you are affected by this change and would like to discuss further options, please contact enterprise@topazlabs.com

Bug fixes & other improvements:

  • UI lag greatly reduced, specifically for long operations.
  • Timeline width now adjusts correctly between different video inputs.
  • Export preferences now save consistently.
  • Added pause/resume controls to right-click menu.
  • Fixed crop settings auto-populating after first digit.
  • Time remaining/elapsed toggle now available for both Previews and Exports.
  • Fixed “Open in Explorer/Finder” button for previews and exports.
  • Disabled automatic generation of TIFF sequences for Previews. These temporary files are no longer in use and the app now reads preview files in the selected Export format.
9 Likes
3

I hope there is a plan to introduce a 2x version of the model before release, otherwise it could be too slow and too resource intensive to be useful compared to other models.

3 Likes
4

Reducing the preview/export list so that only a single line is visible isn’t something I want.

:-1:t2:

1 Like
5

Thanks for the update, but I would like to see:

  1. The new “Rhea” model also compatible with interlaced videos, and thus with the faculty to be used to deinterlaced videos!
  2. The arrival of the long-awaited plug-in for ADOBE PREMIERE PRO!

Thanks!!! :grin:

1 Like
6

something of strange, i tried by chrome and firefox, but i can download windows installer, but if i click on mac installer or dmg give me access denied.

7

Video AI Pro pricing?

8

Looking forward to Video AI Pro
Can’t wait to see it

10

Why is it that when I click to download on a Mac, I’m denied access?

11

So I have a question regarding this new “Pro” license…I see that it is $1099/year for 1 seat…definitely steep compared to the current release. I personally don’t need multi-GPU rendering but I was concerned about the “commercial use” part. Does this mean we can no longer use our standard license to upscale for profit?

3 Likes
12

it is not the 1st time that we see AI models initially support x4 only, Iris was at the same spot when it was initially released for Alpha/Beta (at the time they called it Proteus v4, later on they decided to call it a totally new AI name separated from Proteus), If I am not mistaken I think that was the case also with Nyx v2. initially it was only designed for x4, later on they adjusted it to other multipliers before release.

@tony.topazlabs This makes me wonder, what is the advantage of x4 or technical challenges compared to other like x1 or x2, etc. that makes topaz initially create/support the AI for x4 only as a baseline? for me it sounds more sophisticated to upscale to x4 then let’s say x2 as there are more pixels to fill in. so how come x4 seems to be the chosen baseline they do for upscaling AI then let say x2?
is x4 easier to upscale (ignore performance)? will it produce better quality then x2 in general?
What is the reason behind the magic number x4 as AI development baseline?

1 Like
13

“Limited” commercial use is part of the standard licence, and the limit is revenue based.

3 Likes
14

Got it, I won’t even be close to that $1,000,000 limit. So I guess any profit under that even for commercial use is acceptable based on the way I’m understanding the agreement.

15

Yeah, it’s been not really useful for a long time, it actually makes the video choppier in some cases, I brought this up years ago, but it may have been left on the back burner too long due to other more important issues that cropped up (as in version 4’s unfortunate debut) and was most likely overlooked.

I’m glad you brought this up!

16

Please don’t downgrade the standard version. 2% is still a lot of customers using multiple GPUs, that you would be flipping off. For me, it makes a huge difference in usability. Go ahead, give me another reason not to renew! :wink:

3 Likes
18

I would be up for a toggle, but not as a definite one line no matter what at all times and use cases.

19

Iris is still far beyond Rhea when it comes to preserving fine details. Rhea takes after Proteus by making the output look computer generated and unnatural – unless Rhea is intentionally being trained for acne removal and skin lightening…

irisVrhea-skin,hair
irisVrhea-skin,hair1022×580 20.3 KB

Like Proteus, Rhea does too much denoising on fine details and gives the output a greasy oil-pastel look (this is a 2x upscale from FHD to 4K; Iris has “Recover Details” set to zero just like Rhea). I’m not seeing a use-case for Rhea that isn’t already handled much better (and faster) by Iris 1, 2, 3 or Proteus 3 and 4. If users want to achieve the “Rhea” effect, they can perform a two-pass Proteus-Iris enhancement with the intermediate resolution at 4x scale – and the added bonus that they can adjust the Recover Details parameter.

10 Likes
20

I am impressed. The last version had a lot of issues but i could still load an input after a minute - this one doesn’t load an input at all.

I wasn’t expecting the issues to get even worse moving from Alpha to Beta. Does this mean I won’t be able to run the program at all when it goes into release?

21

That looks very much like the difference between the Low Resolution v1 and Low Resolution v2 models in Gigapixel. V2 is now being pushed as the default but v1 does a much better job of retaining fine detail.

3 Likes
22

Dear Topaz Team,

I hope this message finds you well. I am writing to discuss some concerns and observations regarding the use of multiple GPUs with VEAI.
As a long-time user of VEAI (since version <1.0), I have been an advocate for CLI and multi-GPU support. In fact, I have purchased several licenses to circumvent the multi-GPU issues. With these changes, I am concerned about the future of CLI support.
In previous versions of VEAI, I noticed that the multi-GPU functionality was not working as expected, which led me to use multiple instances of VEAI as a workaround. I am pleased to see that this issue seems to have been addressed in the recent updates.

However, I have some concerns about Standard version losing multi gpu to the PRO license model, which appears Topaz are targeting studios/teams engaged in mass conversions. I am curious to know if this change implies that support Standard version conversion with CPU, integrated GPU (iGPU), and discrete GPU will be discontinued. Users are missing out on maxing out there desktop/laptop for the conversion.

The move to limit usage to a single GPU could be quite restrictive for some users. Given the high cost and computational demands of faster, for newer GPU models. A desktop users 16-32Core CPU, with IGPU and two gaming GPUs a limit of three GPUs for the standard version might be a more balanced approach.

Lastly, could you please provide information on the cost implications of upgrading from a standard to a PRO Multi-GPU license?

Please note I gave up on multi GPU in the VEIA 4 Series after trying for a year, and beta testing under VEIA 3 series.

To drop multi GPU from Standard, your price must also be dropping for that licence :slight_smile:

I look forward to your response and appreciate your attention to these matters.

Best regards,

5 Likes
23

Rhea has some big issues with depth of field preservation. I’ve noticed several scenes where it tries to bring the background, which is purposely not sharp/in-focus into focus and it looks very wrong.

This model needs to be adjusted or perhaps retrained to not deblur what should be blurred.

Also not for nothing, but there’s gotta be a model sooner than later that stabilizes noise. Neat Video 5 can do it to a degree, but obviously could be vastly improved (I notice ghosting at times). It’s been a little over a year since I brought this to Topaz’s attention: Transformer for Video Denoising

Btw seeing these pop-up here n’ there in the processed video:
image

2 Likes