Topaz Video 1.4.0

Hi Everyone,

We have a short but extra sweet release!

The team’s been busy and working hard to get this latest update for you!

I’m pleased to announce our most powerful model release in a while! Starlight Precise 2.5 for local desktop rendering!

See more details below:


1.4.0 Downloads
Please note, 1.4.0 release with local features are Windows only.
Mac version includes cloud access for Precise 2.5.


Improvements and Fixes

Features

  • Starlight Fast Precise 2.5 model
  • Neuroserver

Starlight Precise 2.5 Local


Starlight Precise 2.5 made its debut in our Cloud app, Astra, with much excitement and positivity. In the meantime, we were also working hard to get this wonderful model for local desktop rendering! I’m happy to announce that the team has just done that! The best part, we’re making this model available for everyone that meets the requirements to run it! This makes Starlight Precise 2.5 the 5th variant to the desktop Starlight Model Family. Note, that previous iterations of this model exists on Astra. This model has been specialized for realistic human face and skin textures in addition to other areas of improvement such as material, texture, and text.

About:

  • This model works for Medium to High Quality content whether this is generative, live action or restoration videos.
  • Improves generated content that may have soft, plastic, or slight artificial looks, however the AI generated content should have solid structure and composition for ideal improvements.
  • There is a slight difference between the local version and the cloud version in order to fit on the desktop.
  • Greatly reduces flickering with smaller textures.
  • Recommended to use video inputs at or under 4K.
  • Outputs up to 4K.
  • Model access via the cloud for all subscription models.
  • Cloud Export is supported on both Mac and Windows.
  • Expect long render times for this diffusion model. Appears to have faster processing compared to Mini and HQ.

Technical things to know:

  • Nvidia Only
  • Available for all licenses to run locally! Yes, you read this right!
  • Requires a minimum of 12GB VRAM though 16GB VRAM is recommended.
  • Targeted for GPU compute compatibility of 8.9. This is also the recommendation.
  • This model is included locally, using neuroserver.

This model is less ideal for content with:

  • Shots with heavy motion blur may be rendered as texture.
  • Not intended for very low/poor condition quality media.
  • Interlaced content needs to be de-interlaced first.
  • Not ideal for content below 480p


Additional Updates to Note!

Neuroserver

  • Neuroserver installation happens through the installer with a checkbox. An in-app update is recommended.
  • All diffusion models (Starlight Family) now require Neuroserver: Starlight Mini, Sharp, Fast 2, and HQ.
  • Quality for the Starlight Models is expected to be the same.
  • If for some reason the Neuroserver is not working as expected, reinstall the app to run the “Repair” mode.

Don’t Forget

Is your system compatible?
Click Here

Documentation
Click Here

Encountered an Issue?

  • If you experience any issues, please ensure that your drivers are up to date. Next to a proper restart or an app reinstall, this is usually the first step to ensure you have the correct drivers running for this application.
  • Sometimes a firewall can prevent the app from starting. If this occurs, be sure to add Topaz on the exception list.

Otherwise, you can easily reach out to our fantastic Support Team: Click Here


Have fun with the new update!

11 Likes

>Have fun with the new update!
YAY ! New models are always fun :grinning_face_with_smiling_eyes:

1 Like

Yayyy! I was waiting for SL Precise 2.5 for so long. It’s genuinely so good, and finally an upgrade over SL mini.

  • Shots with heavy motion blur may be rendered as texture.

If a future SL Precise 3 or another SL variant can fix this issue, it would genuinely become the definitive upscale model for me. I know this will be challenging, but I trust the Topaz team.

Keep up the great work.

1 Like


I got this message when trying to run Starlight mini using Topaz Video 1.4.0. I granted Topaz Video full disk access but the message remained.

Will these models ever work with Mac? Is there any development happening for Apple hardware?

2 Likes

Where does Neuroserver install to? I’m hoping it defaults to where I have Topaz which is on a non-OS drive.


Starlight still doesn’t work on AMD GPU.
Problem exists since 1.3.0.

I installed the new version and tried Starlight Precise, but I’m getting an error and it’s not working.

The same thing happens with Starlight Mini as well.
Is there a problem with this?

Is anyone else able to run the program without any issues?

2 Likes

Despite what this message says, full disk access should never be requested by, or given to Topaz apps.

I will NEVER give full disk access to any app by Topaz given how buggy their software is. The error message itself appears to show the wrong location “/Models”.

Andy

1 Like

I’ve just tried Starlight Precise and I found it amazing in details and textures!
It’s working fine on 9950X3D and 5090, but the fps and remaining time is clearly wrong. It’s actually 0.1fps~0.2fps and should be around 17 hours.

image

image
Here’s a comparison of Starlight Sharp and Precise. The images are 599 x 450 crops from the final outputs.
540p Starlight mini → 1080p → Starlight Sharp → 4k:

540p Starlight mini → 1080p → Starlight Precise → 4k:

Precise looks even sharper than Sharp and the skin texture is great!

5 Likes

Hello Topaz team, the time code display format didn’t change when i change to frame numbers, it started since 1.3.0 if i installed fresh (uninstall previous version before install), please fix it, this is crucial for my workflow as i need to cut in/out using frame.

Could you check in the \ProgramData\Topaz Labs LLC\Topaz Video\models\models\ (yes, ‘models’ is created twice), if the model blobs were downloaded correctly?

That’s the sizes I’ve got after downloading the blobs:

Though, I’m yet to re-download the Starlight-HQ blobs (previously, they were incomplete, just 2.8GB got dl-ed). How could I force TV 1.4.0 to re-get the model?

There’s also an issue with the export queue preview, where, after the processing is finished, the preview shows the original video ONLY, even after pressing mouse key on the image.

The ‘Reimport export as source' option, after finishing with Starlight model (probably every one, but I had Fast 2 at the time), shows me black video in the ‘editor’, even though the result video is totally fine, and I can see the frames at the bottom.

It lands right in the TV’s folder.

i think there are something wrong with the model download. In previous starlight mini, the model is downlaoded as ‘model.part1.bin.enc’….now all the starlight models downloaded as ‘blob.xxxxxxxxxxxx…’ which isn’t right.

The ‘blobs’ are for the neuroserver.exe, as it’s in charge now for the Starlight models. :slight_smile:

1 Like

Something is definitely wrong with this version.

Starlight Precise doesn’t work at all for me. I just keep getting an “error” message.

With the other Starlight models, the resolution is completely ignored; it always renders at the minimum resolution (720p).

A complete reinstallation didn’t help at all.

I’m also getting an error with the Precise model as well. The other models seem to work and complete about a minute long test video, but the Precise I keep getting an error on too. Below is my third round of testing on each model. They all finish and process except for Precise. It errors out almost immediately before the processing even really starts.

I got an Error on Precise Model too. It isn’t working.

1 Like

Update: Only Starlight Sharp ignores the selected resolution and renders at the minimum resolution of 720p.

However, Starlight Precision still does not work. An error occurs immediately upon loading the model.

2 Likes

I can also confirm this error only on Starlight Precise. I have an RTX 4090 and I was trying to do a 1 min clip 720x480 and it crashes with an instant error. I kept the resolution to the minimum suggested by the model if that makes any difference. All other models works except the new one. I must add, having it run a full length clip makes no difference. The model crash instantly.

logsForSupport.zip (24.4 KB)

2 Likes