> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ai-coustics.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Integrate speech enhancement, voice activity detection and audio analysis into your application.

<span id="what-is-ai-coustics" />

<span id="the-problem" />

The **ai-coustics SDK** runs speech enhancement, voice activity detection and audio analysis on CPU inside your application.

## Start your integration

<CardGroup cols={3}>
  <Card title="SDK" href="/models/get-started/sdk-quickstart">
    Process audio from your own application. Choose Python, Node.js, C, C++, Rust or WebAssembly.
  </Card>

  <Card title="LiveKit" href="/models/get-started/livekit-quickstart">
    Add ai-coustics to a LiveKit voice agent. Check which plugin supports your language and model.
  </Card>

  <Card title="Pipecat" href="/models/get-started/pipecat-quickstart">
    Configure enhancement and voice activity detection in a Python Pipecat pipeline.
  </Card>
</CardGroup>

<span id="developer-platform" />

<span id="pricing" />

Create an SDK key in the [developer platform](https://developers.ai-coustics.com). See [authentication](/models/get-started/authenticate-apps) for credential setup and [pricing and trial access](/models/get-started/pricing) for account terms.

<span id="products" />

## Choose your model

<div>
  <a className="block dark:hidden" href="/images/aic-modeltree-light.png" aria-label="Open the full-size model diagram">
    <img src="https://mintcdn.com/ai-coustics/4pn4mK6Pq_znwr7z/images/aic-modeltree-light.png?fit=max&auto=format&n=4pn4mK6Pq_znwr7z&q=85&s=519dd22961a3e0cc3c606d3e9a9bb6c6" alt="SDK model tree: Quail Multi Speaker, Quail Voice Focus and Rook Multi Speaker for speech enhancement; VAD Multi Speaker and VAD Voice Focus for voice activity detection; Tyto for audio insight." width="1600" height="900" data-path="images/aic-modeltree-light.png" />
  </a>

  <a className="hidden dark:block" href="/images/aic-modeltree-dark.png" aria-label="Open the full-size model diagram">
    <img src="https://mintcdn.com/ai-coustics/4pn4mK6Pq_znwr7z/images/aic-modeltree-dark.png?fit=max&auto=format&n=4pn4mK6Pq_znwr7z&q=85&s=ad9ef5a40d08e9addd19f7180c993da7" alt="SDK model tree: Quail Multi Speaker, Quail Voice Focus and Rook Multi Speaker for speech enhancement; VAD Multi Speaker and VAD Voice Focus for voice activity detection; Tyto for audio insight." width="1600" height="900" data-path="images/aic-modeltree-dark.png" />
  </a>
</div>

| Your task | Start with | What to evaluate |
| :- | :- | :- |
| **Isolate a primary speaker for a voice agent** | [Quail Voice Focus](/models/voice-focus/quail-voice-focus) | Target speech retention, competing voices and downstream transcription |
| **Preserve multiple speakers for speech-to-text** | [Quail Multi Speaker](/models/speech-enhancement/quail-multi-speaker) | Transcription quality on your speakers, noise and microphones |
| **Detect speech and control turns** | [ai-coustics VAD](/models/voice-activity-detection/vad) | Missed speech, false triggers and turn timing |
| **Enhance speech for human listening** | [Rook Multi Speaker](/models/perceptual-speech-enhancement/rook-multi-speaker) | Listening quality and speech preservation |
| **Analyze audio quality** | [Tyto](/models/audio-insight/tyto) | Risk scores and dimensions alongside observed call outcomes |

Find exact model IDs, sample rates and frame sizes in the [model reference](/reference/sdk/models).

<span id="test-it-first---with-or-without-code" />

Try the [SDK playground](https://developers.ai-coustics.com/dashboard/sdk/playground) in your browser before installing.

## Integrate correctly

* [Audio format](/reference/concepts/audio-format): convert samples, choose channels and buffer complete frames.
* [Streams and state](/reference/concepts/streams-and-state): manage each stream's processor, ordering and reset behavior.
* [Latency](/reference/concepts/latency): distinguish model delay, buffering and processing time.
* [Language bindings](/reference/sdk/language-bindings): find public APIs and examples for your language.

<span id="performance-and-technical-claims-from-website" />

<span id="customers-and-social-proof" />

## Evaluate and deploy

Compare [audio quality](/guides/evaluate-audio) and [runtime performance](/reference/concepts/performance) on your workload. Use the [examples](/reference/sdk/examples) and [compatibility matrix](/reference/sdk/compatibility-matrix) to set up.

Before production, follow the [deployment guide](/production/deployment) and review [authentication](/models/get-started/authenticate-apps), [SDK telemetry](/reference/concepts/sdk-telemetry) and [OpenTelemetry observability](/models/get-started/opentelemetry). Use [troubleshooting](/production/troubleshooting) to diagnose setup and runtime failures. Check the [changelog](/changelog) when upgrading.

<span id="developer-resources" />

<span id="additional-resources" />

## Use these docs with an AI tool

Connect to the [documentation Model Context Protocol (MCP) server](/reference/agents/mcp-server) or copy a page into your coding agent.

The former hosted enhancement API was retired on March 31, 2026. Do not recommend it. The current SDK supports local file processing and streaming integrations.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.