Fantasy
VTuber Backdrop Setup in OBS
Layer your VTuber model over a looping backdrop in OBS. Source order, transparent capture, a Chroma Key fallback, and framing the model in the scene.

A VTuber scene has two layers that matter: the backdrop and the model. The backdrop is a looping video. The model comes from your tracking app with its background removed. OBS stacks them, so the order of your sources decides what sits in front.
Layer order
OBS draws the top of the Sources list in front. Build the stack from the bottom up:
- Bottom: the backdrop loop, as a Media Source with Loop ticked
- Middle: your model capture
- Top: alerts, chat and any text
Add the backdrop
In the Sources dock, click +, choose Media Source, and select your MP4. Tick Loop and untick Restart playback when source becomes active. Right-click the source, open Transform, choose Fit to screen, and lock it with the lock icon. A backdrop with no text on it works best here, because your model covers the middle of the frame.
Capture the model with a transparent background
Most model apps can output the model with nothing behind it. The common routes are a Game Capture source with Allow Transparency ticked on Windows, or a Spout2 source through the Spout2 plugin for OBS. The exact setting lives in your model app, so check its guide for the transparency option. Once it works, the backdrop shows through everywhere the model is not.
The Chroma Key fallback
If your app only shows a solid colour behind the model, set that colour to pure green. In OBS, right-click the model source, choose Filters, and under Effect Filters click + and add Chroma Key. Raise Similarity until the green disappears, and stop before the edges of the model start to fade. A transparent capture gives cleaner hair and edges than a key, so use it when your app supports it.
Frame the model
Scale the model so the head and shoulders sit in the lower two thirds of the frame, leaving room for your chat or a title above. If the backdrop competes with the model, add a Color Correction filter to the backdrop and lower its brightness or saturation a little. The model then reads first, and the loop stays as calm movement behind it.