Audio

Text-to-speech (TTS) and speech-to-text (STT) capabilities for voice-enabled applications.

Prism Reference

Atlas audio wraps Prism's audio API. For detailed documentation including all provider options, see Prism Audio.

Text to Speech

Convert text to spoken audio:

use Atlasphp\Atlas\Atlas;

$response = Atlas::audio()
    ->using('openai', 'tts-1')
    ->withVoice('nova')
    ->fromText('Hello, welcome to our service!')
    ->asAudio();

// Save audio file
Storage::put('welcome.mp3', $response->audio);

Speech to Text

Transcribe audio to text:

$response = Atlas::audio()
    ->using('openai', 'whisper-1')
    ->fromFile('/path/to/audio.mp3')
    ->asText();

echo $response->text;  // Transcribed text

Text-to-Speech Examples

With Voice Selection

$response = Atlas::audio()
    ->using('openai', 'tts-1')
    ->withVoice('onyx')  // Deep, authoritative voice
    ->fromText('Important announcement.')
    ->asAudio();

OpenAI voices: alloy, echo, fable, onyx, nova, shimmer

HD Quality

$response = Atlas::audio()
    ->using('openai', 'tts-1-hd')
    ->withVoice('alloy')
    ->fromText('High-definition audio quality.')
    ->asAudio();

Speech-to-Text Examples

Basic Transcription

$response = Atlas::audio()
    ->using('openai', 'whisper-1')
    ->fromFile($request->file('audio')->path())
    ->asText();

return response()->json([
    'text' => $response->text,
]);

With Language Hint

$response = Atlas::audio()
    ->using('openai', 'whisper-1')
    ->fromFile('/path/to/audio.mp3')
    ->withProviderMeta('openai', ['language' => 'en'])
    ->asText();

Complete Examples

Voice Notification Service

class NotificationService
{
    public function sendVoiceNotification(User $user, string $message): void
    {
        $response = Atlas::audio()
            ->using('openai', 'tts-1')
            ->withVoice('nova')
            ->fromText($message)
            ->asAudio();

        $filename = 'notifications/' . Str::uuid() . '.mp3';
        Storage::put($filename, $response->audio);

        $this->voiceService->call($user->phone, Storage::url($filename));
    }
}

Meeting Transcription

class MeetingService
{
    public function processRecording(string $audioPath): array
    {
        // Transcribe
        $transcription = Atlas::audio()
            ->using('openai', 'whisper-1')
            ->fromFile($audioPath)
            ->asText();

        // Summarize with AI
        $summary = Atlas::agent('summarizer')
            ->chat("Summarize this meeting transcript:\n\n{$transcription->text}");

        return [
            'transcript' => $transcription->text,
            'summary' => $summary->text,
        ];
    }
}

Voice Characteristics

Voice	Description
alloy	Neutral, balanced
echo	Clear, confident
fable	Warm, expressive
onyx	Deep, authoritative
nova	Friendly, natural
shimmer	Clear, energetic

Pipeline Hooks

Audio operations support pipeline middleware for observability:

Pipeline	Trigger
audio.before_audio	Before text-to-speech
audio.after_audio	After text-to-speech
audio.before_text	Before speech-to-text
audio.after_text	After speech-to-text

use Atlasphp\Atlas\Contracts\PipelineContract;

class LogAudioGeneration implements PipelineContract
{
    public function handle(mixed $data, Closure $next): mixed
    {
        $result = $next($data);

        Log::info('Audio generated', [
            'user_id' => $data['metadata']['user_id'] ?? null,
        ]);

        return $result;
    }
}

$registry->register('audio.after_audio', LogAudioGeneration::class);

Next Steps

• Prism Audio — Complete audio reference
• Chat — Combine with chat for voice assistants
• Pipelines — Add observability to audio operations