Amigo AIAmigo AI
Back to Docs

How do I integrate Amigo AI SDK into my iOS app?

The SDK provides three integration levels: drop-in SwiftUI/UIKit views, a per-frame API for custom pipelines, and static image swap. All run on-device via CoreML.

System Requirements

  • iOS 16.0 or later
  • Xcode 15.0 or later
  • Swift 5.9 or later
  • Device with A12 Bionic chip or later

Installation (Swift Package Manager)

In Xcode: File → Add Package Dependencies → enter:

https://github.com/AmigoAIAdmin/amigo_sdk_reference.git

Step 1: Download Models (Free)

ML models are downloaded from CDN on first use — not bundled in your app binary. Only emap.bin (1MB) is bundled. Call this during onboarding or a loading screen:

try await AmigoFaceSwap.downloadModelsIfNeeded { progress in
    print("Downloading models: \(Int(progress * 100))%")
}

This is a free operation — no license charge. If models are already cached, it returns immediately.

Step 2: Initialize ($0.01/session)

try await AmigoFaceSwap.initialize(apiKey: "amigo_sk_your_key")

This validates your API key, downloads models if needed, and pre-warms the CoreML engine so the first frame has no cold-start delay.

Step 3: Enroll a Face

Extract a reusable face embedding from a photo:

let latent = try await AmigoFaceSwap.enrollFace(from: targetPhoto)

The returned FaceLatent is a 512-dimensional embedding. Cache it and reuse it — switch targets at runtime with zero latency. You can also construct from pre-computed server-side embeddings:

let latent = FaceLatent(embedding: myFloatArray) // [Float] with 512 elements

Live Camera — SwiftUI

import AmigoFaceSwapSDK
import SwiftUI

struct FaceSwapView: View {
    let latent: FaceLatent

    var body: some View {
        AmigoLiveCameraView(targetLatent: latent)
            .ignoresSafeArea()
    }
}

To receive each processed frame (for recording or streaming):

AmigoLiveCameraView(targetLatent: latent) { frame in
    myEncoder.encode(frame) // UIImage
}

Live Camera — UIKit

let vc = AmigoLiveViewController(targetLatent: latent)
present(vc, animated: true)

Advanced: AmigoLiveSession

For full control over the camera session:

let session = AmigoLiveSession(targetLatent: latent)
session.delegate = self
view.addSubview(session.previewView)
session.previewView.frame = view.bounds
session.start()

// Switch faces at runtime
session.targetLatent = newLatent

// Background replacement
session.backgroundImage = UIImage(named: "office")

// Toggle face swap on/off
session.isFaceSwapEnabled = false

Per-Frame API (WebRTC / Custom Pipelines)

For WebRTC, video playback, or any custom frame source:

// Call from a serial background queue (e.g., AVFoundation videoDataOutputQueue)
if let output = try AmigoFaceSwap.processFrame(pixelBuffer, using: latent) {
    // render output (CIImage) to your display layer
}
// Returns nil when no face detected — render the original frame

Static Image Swap

let result = try await AmigoFaceSwap.swapFace(
    in: inputImage,
    using: latent,
    lipMode: .innerLips
)
imageView.image = result

LipMode

Control lip blending behavior:

| Mode | Description | |------|-------------| | .none | Source face lips as-is | | .outerLips | Blend outer lip region | | .innerLips | Blend inner lip region (default, most natural) |

Cache Management

// Clear all cached models (forces re-download)
AmigoFaceSwap.clearModelCache()
Getting StartedAPI Reference