Generate audio

Use generateAudio for prompt-to-audio models that are not text-to-speech, such as ElevenLabs music and sound effects.

import Foundation
import SwiftAISDK

@main
struct GenerateAudioExample {
    static func main() async {
        do {
            let provider = try AIProviders.elevenLabs()

            let soundEffects = try provider.soundEffectsModel()
            let click = try await AI.generateAudio(
                "A single soft interface click.",
                using: soundEffects,
                durationSeconds: 0.5,
                format: "mp3_32",
                providerOptions: ["elevenlabs": ["promptInfluence": 0.3]]
            )
            print("Generated sound effect bytes: \(click.audio.count)")

            let music = try provider.musicModel()
            let cue = try await AI.generateAudio(
                "Three seconds of quiet instrumental piano, no vocals.",
                using: music,
                durationSeconds: 3,
                format: "mp3_32",
                providerOptions: ["elevenlabs": ["forceInstrumental": true]]
            )
            print("Generated music bytes: \(cue.audio.count)")
        } catch {
            print("Audio generation failed: \(error)")
        }
    }
}

AudioGenerationRequest carries:

prompt: the natural-language audio prompt.
durationSeconds: optional requested duration.
format: provider output format such as mp3_32 or mp3_44100_128.
seed, providerOptions, extraBody, headers, and abortSignal.

AudioGenerationResult returns generated audio bytes, content type, warnings, provider metadata, request metadata, and response metadata.

ElevenLabs exposes provider.musicModel() for /v1/music and provider.soundEffectsModel() for /v1/sound-generation.