ElevenLabs

Advanced speech-to-text with diarization and audio event tagging

Configuration

'elevenlabs' => [
    'api_key' => env('ELEVENLABS_API_KEY', ''),
    'url' => env('ELEVENLABS_URL', 'https://api.elevenlabs.io/v1/'),
]

Speech-to-Text

ElevenLabs provides speech-to-text through their Scribe model with support for diarization and audio event tagging.

Basic Usage

use Prism\Prism\Facades\Prism;
use Prism\Prism\ValueObjects\Media\Audio;

$audioFile = Audio::fromPath('/path/to/recording.mp3');

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($audioFile)
    ->asText();

echo $response->text;

Provider-Specific Options

Language Detection

Specify the language code for better transcription accuracy:

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($audioFile)
    ->withProviderOptions([
        'language_code' => 'en',
    ])
    ->asText();

Speaker Diarization

ElevenLabs can identify and separate different speakers in the audio:

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($audioFile)
    ->withProviderOptions([
        'diarize' => true,
        'num_speakers' => 2,
    ])
    ->asText();

// Access speaker information
$segments = $response->additionalContent['segments'] ?? [];
foreach ($segments as $segment) {
    echo "Speaker {$segment['speaker']}: {$segment['text']}\n";
}

Audio Event Tagging

Detect non-speech audio events like laughter, applause, or background noise:

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($audioFile)
    ->withProviderOptions([
        'tag_audio_events' => true,
    ])
    ->asText();

// Events are included in the transcription
echo $response->text;
// Example: "Hello [LAUGHTER] how are you? [APPLAUSE]"

Use Cases

Meeting Transcription with Speaker Identification

$meetingAudio = Audio::fromPath('/path/to/meeting.mp3');

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($meetingAudio)
    ->withProviderOptions([
        'diarize' => true,
        'num_speakers' => 4,
        'language_code' => 'en',
        'tag_audio_events' => true,
    ])
    ->asText();

// Process segments with speaker labels
$segments = $response->additionalContent['segments'] ?? [];
foreach ($segments as $segment) {
    echo "[Speaker {$segment['speaker']}] {$segment['text']}\n";
}

Podcast Transcription

$podcastAudio = Audio::fromUrl('https://example.com/podcast.mp3');

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($podcastAudio)
    ->withProviderOptions([
        'diarize' => true,
        'num_speakers' => 2,  // Host and guest
        'tag_audio_events' => true,  // Capture laughter, music, etc.
    ])
    ->asText();

Interview Transcription

$interviewAudio = Audio::fromPath('/path/to/interview.wav');

$response = Prism::audio()
    ->using('elevenlabs', 'scribe_v1')
    ->withInput($interviewAudio)
    ->withProviderOptions([
        'diarize' => true,
        'num_speakers' => 2,
        'language_code' => 'en',
    ])
    ->asText();

// Generate formatted transcript
$segments = $response->additionalContent['segments'] ?? [];
$speakers = ['Interviewer', 'Guest'];

foreach ($segments as $segment) {
    $speakerLabel = $speakers[$segment['speaker'] - 1] ?? "Speaker {$segment['speaker']}";
    echo "{$speakerLabel}: {$segment['text']}\n\n";
}

Audio File Handling

Supported Formats

ElevenLabs Scribe supports various audio formats:

use Prism\Prism\ValueObjects\Media\Audio;

// From local file path
$audio = Audio::fromPath('/path/to/audio.mp3');
$audio = Audio::fromPath('/path/to/audio.wav');
$audio = Audio::fromPath('/path/to/audio.m4a');

// From remote URL
$audio = Audio::fromUrl('https://example.com/recording.mp3');

// From base64 encoded data
$audio = Audio::fromBase64($base64AudioData, 'audio/mpeg');

// From binary content
$audioContent = file_get_contents('/path/to/audio.wav');
$audio = Audio::fromContent($audioContent, 'audio/wav');

Features

• ✅ Speech-to-Text with high accuracy
• ✅ Speaker Diarization (identify multiple speakers)
• ✅ Audio Event Tagging (detect non-speech sounds)
• ✅ Multi-language support
• ❌ Text-to-Speech (not yet implemented)

Best Practices

For Best Diarization Results

1. Ensure clear audio quality
2. Minimize background noise
3. Specify the correct number of speakers
4. Use a sample rate of at least 16kHz

For Accurate Transcription

1. Use the correct language code
2. Ensure good audio quality (clear speech, minimal noise)
3. Use appropriate audio format (WAV or high-quality MP3)
4. For long recordings, consider splitting into segments

Limitations

Text-to-Speech

ElevenLabs text-to-speech is not yet implemented in Prism. Use OpenAI or Groq for TTS functionality.

File Size

Check ElevenLabs documentation for current file size limits when processing audio files.