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

# STT Testing

> Evaluate speech-to-text providers with accuracy metrics

## Overview

Destined Voice includes tools to evaluate Speech-to-Text (STT) providers. Test multiple providers against your audio and analyze accuracy with WER/CER metrics.

## Supported Providers

| Provider          | Model             | Description           |
| ----------------- | ----------------- | --------------------- |
| Deepgram Nova-3   | `deepgram-nova-3` | Latest Deepgram model |
| Deepgram Flux     | `deepgram-flux`   | Real-time optimized   |
| AssemblyAI        | `assemblyai`      | High accuracy         |
| OpenAI Whisper    | `openai`          | Multilingual          |
| Google Speech     | `google`          | Google Cloud STT      |
| Azure Speech      | `azure`           | Microsoft Azure       |
| Amazon Transcribe | `amazon`          | AWS Transcribe        |
| Soniox            | `soniox`          | Low latency           |
| Play.ht           | `playht`          | Specialized model     |

## Transcribing Audio

Send audio to multiple providers:

```typescript theme={null}
const results = await client.sttTesting.transcribeV1SttTranscribePost({
  audioUrl: "https://example.com/audio.wav",
  providers: ["deepgram-nova-3", "assemblyai", "openai"],
});

console.log(results);
// {
//   "deepgram-nova-3": {
//     transcript: "Hello, this is a test.",
//     latency_ms: 450,
//     confidence: 0.98
//   },
//   "assemblyai": {
//     transcript: "Hello, this is a test.",
//     latency_ms: 620,
//     confidence: 0.97
//   },
//   ...
// }
```

## Calculating Accuracy

Compare transcriptions against ground truth:

```typescript theme={null}
const accuracy = await client.sttTesting.calculateAccuracyV1SttCalculateAccuracyPost({
  reference: "Hello, this is a test.",
  hypothesis: "Hello, this is the test.",
});

console.log(accuracy);
// {
//   wer: 0.20,        // Word Error Rate (20%)
//   cer: 0.05,        // Character Error Rate (5%)
//   substitutions: 1,
//   insertions: 0,
//   deletions: 0
// }
```

## Metrics Explained

### Word Error Rate (WER)

Measures word-level accuracy:

```
WER = (Substitutions + Insertions + Deletions) / Total Reference Words
```

* **Lower is better** (0.0 = perfect, 1.0 = completely wrong)
* Industry standard for STT evaluation

### Character Error Rate (CER)

Measures character-level accuracy:

```
CER = (Substitutions + Insertions + Deletions) / Total Reference Characters
```

* More granular than WER
* Useful for detecting minor transcription errors

## Demographic Bias Analysis

Analyze STT accuracy across demographics:

```typescript theme={null}
// Generate test audio with different speakers
const speakers = await client.speakers.listSpeakersV1SpeakersGet({
  limit: 100,
});

// Group by demographic and compare WER
// Enterprise feature - contact sales
```

## Best Practices

<AccordionGroup>
  <Accordion title="Use consistent audio quality">
    Test with audio at 16kHz or higher. Lower quality affects all providers equally.
  </Accordion>

  <Accordion title="Normalize text before comparison">
    Remove punctuation, lowercase text, and expand numbers for fair WER calculation.
  </Accordion>

  <Accordion title="Test multiple accents">
    STT accuracy varies by accent. Test with speakers matching your user base.
  </Accordion>

  <Accordion title="Consider latency vs accuracy">
    Some providers trade accuracy for speed. Choose based on your use case.
  </Accordion>
</AccordionGroup>

## Provider Comparison (Typical Performance)

| Provider        | Avg WER | Avg Latency | Best For      |
| --------------- | ------- | ----------- | ------------- |
| Deepgram Nova-3 | 5-8%    | 400ms       | General use   |
| AssemblyAI      | 4-7%    | 600ms       | High accuracy |
| OpenAI Whisper  | 5-10%   | 800ms       | Multilingual  |
| Google          | 6-10%   | 500ms       | Integration   |
| Azure           | 6-10%   | 550ms       | Enterprise    |

<Note>
  Actual performance varies by audio quality, accent, and domain vocabulary.
</Note>