Cinematic. Superpowers.

Plans
  • Paid
  • Free Trial
Platforms
Social Links

Description

Production-grade AI models for professional use cases: Topaz Video AI focuses solely on completing a few video enhancement tasks really well: deinterlacing, upscaling, and motion interpolation. We've taken five years to craft AI models robust enough for natural results on real-world footage.
‍Topaz Video AI will also take full advantage of your modern workstation, as we partner directly with hardware manufacturers to optimize processing times. (Many of them already use Topaz Video AI to benchmark AI inference.) Own the software and use it for as many projects as you like, right in your existing workflow.

Fewer motion artifacts: Other video upscaling techniques often create a “shimmering” or “flickering” effect from different processing in adjacent frames. TVAI significantly reduces these artifacts.

Recover video details: Inject real details into your videos, derived from the additional information in multiple adjacent frames.

Get more natural results:Traditional upscaling often causes artifacts. We mitigate these by deriving new information from multiple frames.

Trained on videos: Topaz Video AI was trained exclusively on video clips instead of still images.

Video enhancement for production workflows: Topaz Video AI uses your existing hardware to directly process videos instead of offloading work to a remote server. This means that you can preview an unlimited number of AI models and parameters to find the perfect result without worrying about server costs. It's also more cost-effective, faster, and better integrated into your workflow than a web-based video upscaler.

Rating and Reviews

Add Rating and Review
0
(0 Reviews)
5  
0%
4  
0%
3  
0%
2  
0%
1  
0%

Questions & Answers

How do we remove noise from images?

Using Autopilot: Autopilot analyzes the sample image and will:

  • detect strong noise
  • turn on Remove Noise

  • select the Strong model

  • set the slider for Strength and Detail automatically

A green dot appears next to Remove Noise and the sliders when using autopilot settings. If you make manual changes to sharpening click the green arrow to return to Autopilot settings.

Manual Controls: Click the toggle to manually disable/enable Remove Noise. For weaker noise reduction, switch to the Normal model.

  • Normal model is for low-medium noise images.
  • Strong model is best for high noise images.
  • Extreme model is for high noise images with better detail retention and controls.

To fine tune the results, change the model settings by moving the sliders.

  • Strength slider controls the amount of noise removed.
  • Detail slider controls minor sharpening on the entire image, recovering detail after noise is removed.

The sample results show the difference in different models. Original image (left), Normal, Strong, Extreme (right).

Oct. 4, 2023


How do we sharpen images?

Using Autopilot: Autopilot analyzes the sample image and will:

  • detect the subject
  • detect blur
  • turn on Sharpen

  • select the Strong model

  • set the slider for Strength automatically

 A green dot appears next to Sharpen and the sliders when you are using autopilot settings. If you make manual changes to sharpening click the green arrow to return to Autopilot settings.

Subject Selection: Hover over the "Subject" text to view the mask created by the AI model.  To improve the mask manually, click the Subject button below the preview or the Edit Subject button. Here you can select the AI model used for subject detection and manually mask the image yourself. Pick the AI model based on the subject in your image. Change the Feathering slider to soften and slightly extend edges of the mask.To manually mask your image, use the AI brush or simple brush. Switch with the toggle.

  • The AI brush breaks up the image into pieces for fast masking.
  • The basic brush is a circular with size control for easy masking.

Apply your masking changes by clicking the Done button at the bottom.

Manual Controls: Click the toggle to manually disable/enable Sharpen.

By default, the subject is sharpened. To sharpen the entire image, toggle Subject Only.

  • The Strength slider controls the amount of sharpening applied.
  • The Clean slider controls minor denoising. The Strong model does not have this.
  • Standard model is best for low and medium blur.
  • Strong model is best for medium and high blur.
  • Lens Blur is for out-of-focus images.
  • Motion Blur is for streaking in images.

Oct. 4, 2023


How do we recover faces in images?

Using Autopilot: Autopilot analyzes the sample image. It will:

  • detect faces in the image
  • detect face quality
  • turn on Face Recovery
  • set the slider Strength automatically
  • turn on Enhance Resolution and upscale the image 3.6x due to low starting quality                                                                            A green dot appears next to Recover Faces and the sliders when you are using autopilot settings. If you make manual changes to Face Recovery strength or selected faces, you can return to autopilot settings by clicking the reset button.                                 

    Face Selection: Hover over the "Faces" text to view the faces detected and the faces currently selected for improvement.To select/deselect faces for processing, click the Faces button under the miniature preview or the Select Faces button in the Recover Faces filter.

    Click on a face to select it. The preview shows:

  • faces selected for processing are outlined in yellow.
  • deselected faces are outlined in grey.
  • Click Apply to save your selection.

    Click "Select All " to toggle all faces or no faces selected for processing.                                                                                   

    Manual Controls: Click the toggle to manually disable/enable Face Recovery. This can be toggled only if faces are detected in your image.For a weaker result decrease the Strength slider. The default Strength is set to 80 and can be changed in the Preferences > Autopilot menu.                                            

Oct. 4, 2023


how to use Topaz Video AI to upscale a low-quality file to HD?

For low-quality files, we recommend using Artemis Low Quality. If the results are not as expected, switch to Medium Quality or Proteus Fine Tune. For this example, we will be using Artemis Low Quality to reduce compression artifacts, reduce noise, and upscale to HD.

The input file (720x576p) is highly compressed and noisy, the Artemis Low-Quality model effortlessly upscales the file to HD while reducing the compression artifacts and reducing noise.

To enhance and upscale your video using Artemis Low Quality, select HD from the video out menu. This will lock the Enhancement filter.

Next, choose Artemis for the AI Model and choose Low Quality, Noisy for the input video condition.

Generate a preview and use the zoom tool to examine the results.

Oct. 4, 2023