steipete/elevenlabskit

Install

Add https://github.com/steipete/ElevenLabsKit.git in Xcode under File → Add Package Dependencies, or add the package and product to your Package.swift:

dependencies: [
    .package(url: "https://github.com/steipete/ElevenLabsKit.git", from: "0.1.1"),
],
targets: [
    .target(name: "YourApp", dependencies: ["ElevenLabsKit"]),
]

The current source tree requires Swift 6.3, iOS 17 or later, or macOS 15 or later.

Quick start

Create a client, stream PCM audio, and play chunks as they arrive:

import ElevenLabsKit

let client = ElevenLabsTTSClient(apiKey: "<api-key>")
let request = ElevenLabsTTSRequest(
    text: "Hello from ElevenLabsKit",
    modelId: "eleven_v3",
    outputFormat: "pcm_44100")
let stream = client.streamSynthesize(voiceId: "<voice-id>", request: request)
let result = await PCMStreamingAudioPlayer.shared.play(stream: stream, sampleRate: 44_100)

The request needs an ElevenLabs API key and a voice ID from your account. result.finished reports whether playback completed; starting another playback session interrupts the active one.

Synthesis

Use synthesize when you need the complete audio payload instead of a stream:

let data = try await client.synthesize(voiceId: "<voice-id>", request: request)

List the voices available to the account with try await client.listVoices(). synthesize retries rate-limit, server, and retryable transport failures up to three attempts; synthesizeWithHardTimeout adds an overall cancellation deadline.

Playback

The package provides two shared players:

| Format | Player | Backend | | --- | --- | --- | | MP3 | StreamingAudioPlayer.shared | Audio Queue Services | | PCM | PCMStreamingAudioPlayer.shared | AVAudioEngine and AVAudioPlayerNode |

Choose an output format that matches the player. For PCM formats, TalkTTSValidation.pcmSampleRate(from:) extracts the rate from values such as pcm_44100:

let sampleRate = TalkTTSValidation.pcmSampleRate(from: request.outputFormat) ?? 44_100
let result = await PCMStreamingAudioPlayer.shared.play(stream: stream, sampleRate: sampleRate)

Call stop() on either player to interrupt playback and receive its last timestamp when available.

Request validation

ElevenLabsTTSRequest supports model, output format, speed, stability, similarity, style, speaker boost, seed, normalization, language, and streaming latency settings. TalkTTSValidation and the static helpers on ElevenLabsTTSClient validate common inputs before a request is sent:

• Speed accepts 0.7...1.2, including conversion from words per minute.
• Stability accepts 0...1; eleven_v3 accepts 0, 0.5, or 1.
• Similarity and style accept 0...1.
• Streaming latency tier accepts 0...4.
• Language accepts two-letter codes, and normalization accepts auto, on, or off.

Examples

ElevenLabsKitExample is a macOS SwiftUI app for comparing streaming and fetched playback, selecting voices, and adjusting request parameters. Open its Package.swift in Xcode and run the ElevenLabsKitExample scheme.

ElevenLabsKitCLI exercises the same APIs from a terminal:

cd Examples/ElevenLabsKitCLI
swift run ElevenLabsKitCLI --help

Speech and voice-list commands require ELEVENLABS_API_KEY or --api-key.

Development

swift test
swiftformat Sources Tests Examples --lint
swiftlint lint --strict --config .swiftlint.yml
(cd Examples/ElevenLabsKitExample && swift build)
(cd Examples/ElevenLabsKitCLI && swift build)

License

ElevenLabsKit is available under the MIT License.

Package Metadata

Repository: steipete/elevenlabskit

Default branch: main

README: README.md