How Video-Conferencing Apps Add Virtual Backgrounds With an SDK

[navigation]

TL;DR

• A virtual background feature keeps cameras on: it hides the user's real surroundings, which removes the "I can't turn my camera on right now" objection that lowers engagement in conferencing products.
• Banuba's background subtraction is powered by a neural network with detection accuracy better than most video-conferencing and streaming apps, so the cut-out edge holds up in real meetings, not just demos.
• Banuba's virtual webcam features (Background Remover, Face Filters, Face Touch Up) run on devices with a 1280x720 camera at a minimum of 30 FPS, which is the frame rate video calls actually run at.
• Processing is fully on-device, so no video frame leaves the user's machine, which matters for corporate and regulated conferencing buyers.
• On the web build, Banuba's core SDK ships as a roughly 12 Mb WASM binary that compresses to about 2.5 to 3.5 Mb over the wire, and the exact footprint depends on the feature set you enable.
• Banuba Face AR SDK covers iOS, Android, Web, Windows, macOS, Unity, Flutter, and React Native from one integration, so a conferencing app on several platforms builds the feature once.

Why do virtual backgrounds matter for video-conferencing apps?

Camera enablement is the metric conferencing products live or die on. When people are reluctant to show their room, they leave the camera off, and a call with black tiles is a worse product than one with faces. A virtual background removes the reason to hide: the user picks an image, a blur, or a branded scene, and their actual surroundings disappear. That protects privacy, sets a professional tone for corporate calls, and quietly raises the share of participants who keep the camera on.

Building this in-house means training and maintaining a person-segmentation model, tuning it for hair, glasses, and low light, and keeping it fast enough for live video. Most conferencing teams do not have computer-vision headcount to spare, and the feature is table stakes rather than a differentiator, so the build-versus-buy math favors an SDK.

Banuba's virtual backgrounds example

How does Banuba's background subtraction solve it?

Banuba's background subtraction is a neural-network filter that separates the person from the scene and replaces everything behind them with a still image, a GIF, or a video. Banuba's documentation states the network's detection accuracy is better than that of most video-conferencing and streaming apps, which is the part that decides whether the edge around hair and shoulders looks clean or ragged on a real call.

Two numbers matter for a conferencing decision-maker. First, Banuba's virtual webcam features, including Background Remover, run at a minimum of 30 FPS on devices with a 1280x720 camera, so the background holds steady at the frame rate calls are streamed at. Second, everything runs on-device: no frame is sent to a server for processing, so background subtraction adds a real privacy guarantee rather than a new data-handling liability. For group calls, Banuba Face AR SDK tracks multiple faces at once, limited only by the device's performance.

What does the integration take to ship?

The SDK distributes as a native package per platform, and the feature is configured rather than coded from scratch. Banuba's Face AR SDK documentation covers setup, and the quickstart repositories for Android, iOS, and Web give you a running camera pipeline to drop the background effect into.

One planning note for web builds: Banuba's core SDK is a roughly 12 Mb WASM binary that compresses to about 2.5 to 3.5 Mb with gzip or brotli, and the total footprint depends on which features you turn on. On web especially, budget for that in your page-load and bandwidth targets. Because the same Face AR SDK covers iOS, Android, Web, Windows, macOS, Unity, Flutter, and React Native, a conferencing app that runs on several platforms integrates the background feature once instead of maintaining a separate implementation per OS.

Does it work with a live-streaming stack?

Yes. Virtual backgrounds are not limited to boxed video-calling apps: a live-streaming app built with Amazon IVS and Banuba Face AR SDK can apply background subtraction and filters to the broadcaster's feed before it goes out, as shown in Banuba's Amazon IVS live-streaming integration guide. The same on-device pipeline that powers a private meeting powers a public broadcast, so a product that spans conferencing and streaming reuses one integration.

Get started

Try Banuba's background subtraction in your conferencing app and see how camera enablement moves once users can hide their room. Get a demo or start a free trial to test it on your own footage.