# STT Providers Overview

> Compare Speech-to-Text providers for accuracy, latency, language support, and cost to choose the right one for your voice agent.

Source: https://edesy.in/docs/voice-agent/stt

Speech-to-Text (STT) converts the caller's audio into text that the LLM can understand. Choosing the right STT provider significantly impacts accuracy, latency, and cost.

## Supported Providers

| Provider | Model | Latency | Languages | Best For |
|----------|-------|---------|-----------|----------|
| [Deepgram](https://edesy.in/docs/voice-agent/stt/deepgram) | Nova-3 | ⚡ Fastest | 35+ | Production voice agents |
| [Google](https://edesy.in/docs/voice-agent/stt/google) | Chirp 2 | 🚀 Fast | 125+ | Indic languages |
| [Azure](https://edesy.in/docs/voice-agent/stt/azure) | Neural | 🚀 Fast | 100+ | Enterprise |
| [ElevenLabs](https://edesy.in/docs/voice-agent/stt/elevenlabs) | Scribe | 🚀 Fast | 99+ | Regional languages |
| [AssemblyAI](https://edesy.in/docs/voice-agent/stt/assemblyai) | Universal-2 | 🚀 Fast | 50+ | Accuracy-focused |
| [OpenAI](https://edesy.in/docs/voice-agent/stt/whisper) | Whisper | 🐢 Moderate | 100+ | Multilingual |

## Quick Comparison

```
Latency (Time to First Partial, lower is better):
──────────────────────────────────────────────────────────────────

Deepgram Nova-3   ████░░░░░░░░░░░░░░░░░░░░░░░░░░░░  80ms
Google Chirp      ████████░░░░░░░░░░░░░░░░░░░░░░░░  120ms
Azure Neural      ████████░░░░░░░░░░░░░░░░░░░░░░░░  130ms
ElevenLabs        ██████████░░░░░░░░░░░░░░░░░░░░░░  150ms
AssemblyAI        ████████████░░░░░░░░░░░░░░░░░░░░  180ms
OpenAI Whisper    ████████████████████████████████  300ms

                  0ms              150ms            300ms
```

## Choosing the Right Provider

### For Low Latency (Recommended)

**Deepgram Nova-3**

- Fastest time-to-first-partial (~80ms)
- Excellent for real-time voice agents
- Best endpointing (detects speech completion)
- Smart formatting included

```json
{
  "sttProvider": "deepgram",
  "sttModel": "nova-3"
}
```

### For Indic Languages

**Google Chirp 2**

- Best accuracy for Hindi, Tamil, Telugu, Bengali
- 125+ languages supported
- Chirp 2 is optimized for telephony

```json
{
  "sttProvider": "google",
  "sttModel": "chirp_2"
}
```

### For Regional Languages

**ElevenLabs Scribe**

- Excellent for Assamese, Odia, Punjabi
- 99 languages with good regional coverage
- Competitive pricing

```json
{
  "sttProvider": "elevenlabs"
}
```

### For Enterprise

**Azure Neural**

- Enterprise SLAs and compliance
- Custom speech models available
- Global deployment options

```json
{
  "sttProvider": "azure"
}
```

## Streaming vs Batch

All our STT integrations use **streaming** for real-time voice agents:

```
Batch STT (not suitable for voice):
────────────────────────────────────
User speaks for 5 seconds → Wait → Get full transcript

Streaming STT (what we use):
────────────────────────────────────
User speaks → Interim results every 100ms → Final transcript
                    │
                    └── LLM can start preparing response
```

### Interim Results

Interim results allow faster response preparation:

```go
// STT emits interim results as user speaks
type TranscriptEvent struct {
    Text      string
    IsFinal   bool
    Stability float32  // 0.0-1.0, higher = more stable
}

// Example stream for "What is my order status?"
// t=100ms: {Text: "What", IsFinal: false, Stability: 0.8}
// t=200ms: {Text: "What is", IsFinal: false, Stability: 0.85}
// t=300ms: {Text: "What is my", IsFinal: false, Stability: 0.9}
// t=500ms: {Text: "What is my order", IsFinal: false, Stability: 0.92}
// t=700ms: {Text: "What is my order status", IsFinal: true, Stability: 1.0}
```

## Endpointing

Endpointing detects when the user has finished speaking:

| Provider | Endpointing | Configurable |
|----------|-------------|--------------|
| Deepgram | ⭐⭐⭐⭐⭐ Smart | Yes |
| Google | ⭐⭐⭐⭐ Good | Yes |
| Azure | ⭐⭐⭐⭐ Good | Yes |
| ElevenLabs | ⭐⭐⭐ Basic | Limited |
| AssemblyAI | ⭐⭐⭐⭐ Good | Yes |

### Deepgram Endpointing Configuration

```json
{
  "sttProvider": "deepgram",
  "sttConfig": {
    "endpointing": 300,
    "utterance_end_ms": 1000,
    "interim_results": true
  }
}
```

| Parameter | Default | Description |
|-----------|---------|-------------|
| `endpointing` | 300 | Silence duration (ms) to trigger is_final |
| `utterance_end_ms` | 1000 | Maximum wait for speech completion |
| `interim_results` | true | Enable real-time partial transcripts |

## Language Support Matrix

| Language | Deepgram | Google | Azure | ElevenLabs | AssemblyAI |
|----------|----------|--------|-------|------------|------------|
| English (US) | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| Hindi | ⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ |
| Spanish | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ |
| French | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| German | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ |
| Tamil | ⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ |
| Telugu | ⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ |
| Bengali | ⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐ |
| Assamese | ❌ | ⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐⭐ | ❌ |
| Japanese | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ |

## Cost Comparison

| Provider | Cost per Minute | Monthly (1M minutes) |
|----------|-----------------|---------------------|
| Deepgram Nova-3 | $0.0043 | $4,300 |
| Google Chirp | $0.016 | $16,000 |
| Azure Neural | $0.016 | $16,000 |
| ElevenLabs | $0.007 | $7,000 |
| AssemblyAI | $0.0055 | $5,500 |
| OpenAI Whisper | $0.006 | $6,000 |

## Audio Requirements

| Parameter | Requirement | Notes |
|-----------|-------------|-------|
| Sample Rate | 8000 Hz | Telephony standard (μ-law) |
| Channels | Mono | Single channel |
| Bit Depth | 16-bit | Linear PCM |
| Encoding | Linear16 or μ-law | Provider dependent |

### Audio Conversion

```go
// Telephony sends μ-law, STT may need Linear16
func mulawToLinear16(mulaw []byte) []int16 {
    linear := make([]int16, len(mulaw))
    for i, sample := range mulaw {
        linear[i] = mulawToLinearSample(sample)
    }
    return linear
}
```

## Error Handling

Handle STT failures gracefully:

```go
func handleSTTError(err error) {
    switch {
    case errors.Is(err, ErrAudioTooQuiet):
        // Ask user to speak louder
        tts.Speak("I'm having trouble hearing you. Could you speak a bit louder?")

    case errors.Is(err, ErrConnectionLost):
        // Reconnect automatically
        stt.Reconnect()

    case errors.Is(err, ErrRateLimited):
        // Use fallback provider
        stt.SwitchProvider("fallback")
    }
}
```

## Next Steps

- [Deepgram Configuration](https://edesy.in/docs/voice-agent/stt/deepgram) - Fastest STT for voice
- [Google Chirp](https://edesy.in/docs/voice-agent/stt/google) - Best for Indic languages
- [Azure Speech](https://edesy.in/docs/voice-agent/stt/azure) - Enterprise deployment
- [ElevenLabs Scribe](https://edesy.in/docs/voice-agent/stt/elevenlabs) - Regional languages