vLLM

¶Audio Transcription with vLLM

This example demonstrates how to transcribe audio files into text using the transcription API of vLLM through the PhpMistral library.

The transcription process sends an audio file (e.g., .mp3, .wav) to a model like Whisper and retrieves the corresponding text output.

Note

This feature is available in vLLM servers configured with a transcription model (such as OpenAI's Whisper family).

¶Example Code

use Partitech\PhpMistral\Clients\Vllm\VllmClient;
use Partitech\PhpMistral\MistralClientException;

$apiKey = getenv('VLLM_API_KEY');   // Your vLLM token
$urlApiTranscription = getenv('VLLM_API_TRANSCRIPTION_URL'); // vLLM transcription endpoint
$transcriptionModel  = getenv('VLLM_API_TRANSCRIPTION_MODEL'); // e.g., 'openai/whisper-small'

$client = new VllmClient(apiKey: (string) $apiKey, url: $urlApiTranscription);

// Path to the audio file to transcribe
$path = realpath("./../../medias/mit.mp3");

try {
    $response = $client->transcription(
        path: $path,
        model: $transcriptionModel
    );
    print_r($response);
} catch (MistralClientException $e) {
    echo $e->getMessage();
    exit(1);
}

¶Example Output

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files.

¶Parameters

¶Supported Models

This endpoint works with transcription models deployed on your vLLM server. Example models include:

• openai/whisper-small
• openai/whisper-medium
• openai/whisper-large

Refer to the vLLM transcription models documentation for a full list.

Tip

The Whisper models from OpenAI are trained to handle multiple languages and various audio formats.

¶Use Cases

• Transcribe interviews, meetings, or lectures into text.
• Process podcasts or voice notes for documentation or analysis.
• Enable speech-to-text functionalities in your applications.

Caution

Ensure that the vLLM server is configured with transcription capabilities and supports the chosen model.