How to Implement a Beauty API in Your App

A beauty API is a documented set of methods that apply face modification and makeup rendering to a live camera feed, a still image, or a prerecorded video. Instead of writing shaders, segmentation models, and a face tracker, a developer calls a named feature and passes it parameters.

The distinction that matters for planning: this is not a REST endpoint. Banuba's beauty modules ship inside Banuba Face AR SDK and execute on the device. There is no upload step, no round trip, and no per-request latency, which also removes the privacy question that comes up in almost every beauty and try-on review, because camera frames never leave the handset.

 AR Beauty and Makeup API overview

Two capability groups sit behind the same face filter. Makeup covers the products a shopper would buy: lipstick, eyeshadow, foundation, blush, brows. Beautification covers how the person wants to look on camera: smoother skin, whiter teeth, softer facial proportions, brighter eyes. Banuba's documentation describes how to combine both in a single effect, which is what most beauty apps end up needing.

Why Would a Developer Integrate a Beauty API Instead of Building It?

Building the same functionality in-house means owning a face tracker, per feature segmentation, a rendering pipeline, and a color model that behaves correctly across skin tones and lighting. It also means maintaining that stack on every OS release.

An SDK-based beauty API changes the shape of the work:

• Effects become configuration files rather than handwritten code, so behavior stays consistent between effects.
• Feature parameters are public and documented, so product teams can adjust a look without a new build of the renderer.
• One integration covers iOS, Android, Web, Flutter, React Native, Unity and desktop.
• Time to market collapses from a research project to an integration task.

For engineering leads evaluating this, the decision usually comes down to two questions: does the tracker hold up in real conditions, and is the documentation good enough that a senior engineer can get a demo running the same week. Banuba publishes platform-specific quickstarts and sample repositories for exactly that reason.

Banuba's beauty makeup effects example 

How Does Banuba's Tracking Engine Support Makeup Rendering?

Makeup rendering is only as good as the face model underneath it. Banuba's engine reconstructs the 3D head model directly from the camera feed instead of inferring 2D landmarks and lifting them into 3D. Skipping that intermediate conversion is what keeps the mesh stable when conditions get difficult.

In practice, that produces 68 facial anchor points and a 3D face mesh with up to 3,308 vertices, with tracking that stays locked across the full -90° to +90° head-angle range and tolerates up to 70% facial occlusion, the kind of coverage you get when a hand, a microphone, or a mask crosses the face.

That stability is the difference between lipstick that stays on the lips during a live stream and lipstick that slides off the moment someone turns their head. Color behavior matters just as much as geometry, and there is a good breakdown of why color accuracy decides whether shoppers trust an AR try-on if you are weighing rendering quality against tracking specs.

Which Makeup Features Can You Implement?

Banuba's current implementation path is makeup prefabs. A prefab represents a specific region of the face and is configured through a small set of parameters, typically color, finish, and coverage. The older Virtual Makeup API and Face Beauty API are both marked deprecated in Banuba's documentation, so new work should start from prefabs.

There are 20 documented makeup prefabs. Grouped by area:

Base and skin

Makeup base, eyebags, foundation, concealer, contour, and highlighter. Foundation and concealer take a color plus a finish (for example, natural) and a coverage level, so a brand can map its real shade range onto the try-on.

Lips

Lipstick, lip liner, lip shine, lip gloss, lip chameleon, and lip glitter. The finish parameter is what separates a matte from a shimmer product, and it is set per prefab rather than baked into a texture.

Eyes and brows

Eyeshadow, eyeliner, eyelashes, eyebrows, eyelid gloss, eyelid chameleon, and eyelid glitter. Eyeshadow accepts an array, so layered looks are described in configuration rather than assembled from flattened textures.

Cheeks

Blush, with the same color, finish, and coverage model as the rest of the set.

All of them can be combined in a single effect definition, which is how a "full look" is assembled from individual product SKUs.

Which Beautification Features Come With It?

Beautification is a separate module set from makeup, and most beauty and streaming apps ship both:

• Skin. Softening and evening out skin without flattening its texture.
• Teeth whitening. Intensity is set on a 0 to 1 scale.
• Face morphing. Adjusts facial proportions by moving vertices on the 3D face model.
• Eye bags removal. Targeted retouching under the eyes.
• Eyes beautification. Including iris color changes for contact lens try-on.
• Hair coloring. Virtual hair shade application.
• Photo filters (LUTs). Color grading applied over the whole frame.
• Background separation. Replacing or blurring what is behind the user.

Live-streaming and video-chat products tend to lean on this set more heavily than on makeup, because the goal is camera confidence rather than product discovery. Natural-looking touch-up is also what a SaaS fandom platform that reached 1 million monthly active users in two years cited when it chose Banuba for its live chat and streaming features, alongside face filters and 3D masks.

How Do You Combine Features Into an Effect?

Each feature is a visual modification of the face or one region of it. A set of features grouped together becomes an effect:

• Effect 1 might contain only a lipstick prefab.
• Effect 2 might contain lipstick, eyeshadow, and face morphing.

Custom textures are supplied by placing the corresponding file into the effect's folder. Supported texture formats include .png and .ktx.

For an e-commerce catalog, this maps cleanly: one effect per product, or one effect per curated look, with the parameters driven by the same color data already stored in the product database. For a selfie or streaming app, effects are usually presets a user can toggle.

What Does the Integration Path Look Like?

The realistic sequence for a first working build:

1. Pick the platform target. Banuba Face AR SDK supports iOS, Android, Web, Flutter, React Native, Unity, and desktop, so decide whether one integration covers the whole roadmap.
2. Run a quickstart. Start from the platform-specific basic integration guides for iOS and Android.
3. Clone a beauty sample. Banuba maintains dedicated beauty repositories for Android, iOS, and web.
4. Build the effect. Define the prefabs, set color, finish and coverage, and drop in any custom textures.
5. Test under stress. Extreme angles, partial occlusion, low light, and older mid-range devices, because that is where 2D-to-3D pipelines usually break first.

There are no code snippets in this guide on purpose. The documentation and the sample repositories are the current source of truth, and they change faster than a blog post does.

From Scratch or SDK: Which Path Fits?

Building from scratch is defensible when face tracking itself is the product. If makeup and beautification are features inside a larger app, the SDK path removes an entire research workstream.

Conclusion

Implementing a beauty API is now an integration decision rather than a computer vision project. The hard parts—a stable 3D face model, per-feature segmentation, and color behavior that survives real lighting—are already solved inside Banuba Face AR SDK. What is left for your team is choosing which of the 20 makeup prefabs and the beautification modules your product actually needs, wiring them to your catalog or your presets, and testing on the devices your users hold.

Start with a quickstart and a beauty sample repository. If the demo runs on a mid-range device the same week, the rest is product work.