Quick Start
The few steps to take to add the Rapidly SDK to your app.
The Rapidly SDK works across Python, Swift, Kotlin, and C / C++ bindings on Windows, Linux, macOS, iOS, and Android. To get started:
1. Choose your binding
Pick the binding that matches your stack. Each binding page has the install command and a working first example.
- Python for servers, ML pipelines, and notebooks.
- Swift for iOS and macOS apps via Swift Package Manager.
- Kotlin for Android apps via Maven Central (
io.rapidly:rapidly-sdk:1.0). - C / C++ for desktop, server, embedded, or any native integration.
See Overview for the full per-platform breakdown.
2. Pick a model
The SDK ships with two model families (speech denoise, speech denoise + dereverb) at four latencies (11, 21, 32, 96 ms), plus a compact micro variant. Choose by latency budget and CPU headroom. See Models for the catalog and picking guidance.
3. Request a license
Without a license, the SDK runs in watermarked demo mode. To unlock clean output, contact us for a license key. Licenses are verified offline, no network required. See Pricing for licensing options.
Next steps
Once you've picked a binding and started integrating, the Guides cover the architectural patterns: when to clean captured audio, when to clean received audio, when to use multiple engines, and parallel processing for cloud workloads.