Vubs and Media

What is a vub?
What are the vub quality levels available in DeepEditor and how should I use them?
How long does each step in the vub creation process take?
How long does it take for a vub to be generated?
Does the audio in the Source media matter?
What is the difference between using Video Driving Data and Audio Driving Data?
How do I stop a vub from generating?
I am getting image artifacts/glitches, what should I do?

What is a vub?

A "vub" refers to a Visual Dub or video modification created using generative AI technology. It is the output you get when you combine Source Media and Driving Data. 

Elements like dialogue, facial expressions, or even whole performances can be altered or substituted while preserving a high degree of realism.

What are the vub quality levels available in DeepEditor and how should I use them?

Please read DeepEditor Vub Quality Levels for full details of each vub quality and how to use it.

How long does each step in the vub creation process take?

• Uploading Source, Driving Data and Training Data (optional) material: On average it takes less than a minute. The length and size of your files may affect this time. 
• Generate a vub: Generating a vub can take between 1 and 12 hours, depending on the vub quality you selected, the length of the media and whether you have added training data. If you haven't received a notification e-mail on completion of a vub and things have taken longer than a day, check the queue to see if the vub failed. See Common Issues and Solutions for more information.
• Download/Review your new file: Your vub will reach your downloads folder in less than a minute for review or conform.
• Render (in the Refinement tool): Rendering a change to your vub can vary based on the amount of changes and length of the clip. On average, rendering will take about 20 minutes.

How long does it take for a vub to be generated?

This depends on a number of factors, such as the resolution and length of your media and whether you included training data. Here are the average times it currently takes to generate a vub:

Does the audio in the Source media matter?

No. The audio in the Source media will not affect the generated vub. Only the audio in the Driving Data will affect your vub.

What is the difference between using Video Driving Data and Audio Driving Data?

If you upload a video file as driving data and it contains both video and audio, we will not process both video and audio and pick the best source to generate the new performance.

If video is uploaded, we will use the performance of the same actor in the video, trying to "copy and paste" to preserve as much of that video performance, and transfer that to the source media. The video will drive the new performance, while the audio will be the audio played in the final vub.

If you upload audio as Driving Data, we'll use that audio to generate the performance.

How do I stop a vub from generating?

Currently, there is no way to stop the process, once you've clicked “Generate Vub”.

I am getting image artifacts/glitches, what should I do?

1. In the refinement tool, review the blendshapes and interpolation you have applied. Lower the interpolation for the affected frames as these glitches happen when the new expression is very different from an expression the AI has seen of the character before.

2. If the artifacts appeared in a Draft Quality Vub, try running a Standard Quality Vub, so you can add Training Data. Adding Training Data can often solve this issue.

Note that we are also working on improvements to our technology that will minimize the instances of these issues appearing. This will be deployed in the coming weeks.

Example of image artifacts / glitches:

For any other issues, please contact us at support@flawless.app.