XangleCs documentation

As of November 2024, Flowframes (abandoned project) has been replace with the much more powerful Topaz Video AI. You’ll need Xangle Camera Server version 24.11.20 (at minimum) to use this new module.

How to use?

1. Download a trial from https://www.topazlabs.com/topaz-video-ai - We made all of our tests on version 5.3. Be careful if you’re using version 5.5 as users reported multiple issues with this one
2. Install Topaz Video AI at the default location
3. Take one reference shot (calibrated)
4. Drop the reference dataset to Topaz Video AI using the default file names (0001.jpg, 0002.jpg, etc)
5. Feed the dataset to Topaz Video AI using the default file names (0001.jpg, 0002.jpg, etc)
6. From the Adjustements panel, set your Frame Interpolation (4x for example)
7. Set the Codec settings to
   • output type = Image Sequence
   • file type = jpg
8. Click on Process / show export command, and copy the command. It should look like this:
9. Paste this string into Xangle under Dashboard / Output Settings / Topaz ffmpeg command

1. To make the interpolation go faster, select a subrange of cameras to interpolate

Tips

• To playback a subset of interpolated images as well as the original (non-interpolated) images, use the Custom-playback-sequence (Dashboard / Output Settings). As an example, 1,2,3,4,i1,i2,i3,i4 will play the first 4 cameras followed by the first 4 interpolated frames.
• Keep your subject small, and in the center of the frame. Interpolation doesn’t work well with subjects that go out of the frame (not enough information to guess what to do)
• You can play with various ramping effects using the Custom Playback Sequence.

Optimizations

• Keep the image size as small as possible (1080p is 4x faster than 4k)
• Apply the interpolation only on a subset of cameras
• Upgrade your graphic card. NVIDIA 4090 @ 24GB is amazing for fast interpolation.

Limitations:

• This doesn’t work in full 360 yet (last to first frame is not interpolated)
• A fast NVIDIA graphic card is required for this. Test your interpolation directly into Topaz before running it into Xangle. You’re going to see right away if your computer can handle this.

We’re very likely to discontinue this integration in 2026 as it’s already quite outdated compared to what can be accomplished with gaussian splat.