ez-php / ai-media

2.7.1 2026-10-30 15:41 UTC

Requires

• php: ^8.5
• ez-php/contracts: ^2.7.0
• ez-php/http-client: ^2.7.0

Requires (Dev)

• ez-php/docker: ^2.7.0
• friendsofphp/php-cs-fixer: ^3.94
• phpstan/phpstan: ^2.1
• phpstan/phpstan-deprecation-rules: ^2.0
• phpstan/phpstan-strict-rules: ^2.0
• phpunit/phpunit: ^13.0

MIT

• Andreas Uretschnig

• php
• ez-php
• ai-media

Published on Packagist.org on 2026-10-30 15:49 UTC

README

Image generation, audio transcription, and text-to-speech via OpenAI/Gemini drivers over ez-php/http-client.

Installation

composer require ez-php/ai-media

Usage

use EzPhp\AiMedia\AiMedia;
use EzPhp\AiMedia\Request\ImageGenerationRequest;
use EzPhp\AiMedia\Request\SpeechRequest;
use EzPhp\AiMedia\Request\TranscriptionRequest;

// Image generation
$images = AiMedia::image(ImageGenerationRequest::make('a red bicycle on a beach'));
$url = $images->first()->url();

// Transcription
$audioBytes = file_get_contents('recording.mp3');
$transcript = AiMedia::transcribe(TranscriptionRequest::make($audioBytes, 'recording.mp3', 'audio/mpeg'));
echo $transcript->text();

// Text-to-speech
$speech = AiMedia::speech(SpeechRequest::make('Hello there.'));
file_put_contents('hello.mp3', $speech->audio());

Request options

Each request is immutable; the with*() methods return a modified copy. withModel() overrides the driver's configured default model for one call.

Request	Option	Default	Notes
ImageGenerationRequest	withCount(int)	1	Images to generate (OpenAI n, Gemini sampleCount)
	withSize(string)	'1024x1024'	OpenAI only; Gemini ignores it
	withModel(string)	config default
SpeechRequest	withVoice(string)	'alloy'	OpenAI voice name, or a Gemini prebuilt voice name (e.g. 'Kore')
	withFormat(string)	'mp3'	OpenAI response_format (mp3, opus, aac, flac, wav, pcm); Gemini always returns raw PCM (audio/pcm)
	withModel(string)	config default
TranscriptionRequest	withLanguage(string)	null (auto-detect)	ISO-639-1 hint sent to OpenAI; Gemini only echoes it back in the response
	withModel(string)	config default

$images = AiMedia::image(
    ImageGenerationRequest::make('a red bicycle on a beach')
        ->withCount(2)
        ->withSize('1792x1024')
        ->withModel('dall-e-3'),
);

$speech = AiMedia::speech(
    SpeechRequest::make('Hallo zusammen.')->withVoice('nova')->withFormat('wav'),
);

$transcript = AiMedia::transcribe(
    TranscriptionRequest::make($audioBytes, 'recording.mp3', 'audio/mpeg')->withLanguage('de'),
);

Register EzPhp\AiMedia\AiMediaServiceProvider and configure config/ai_media.php:

return [
    'image_driver' => 'openai',
    'transcription_driver' => 'openai',
    'speech_driver' => 'openai',
    'openai' => ['api_key' => getenv('OPENAI_API_KEY') ?: null],
    'gemini' => ['api_key' => getenv('GEMINI_API_KEY') ?: null],
];

image_driver, transcription_driver, and speech_driver are selected independently (openai, gemini, or null — the default) since a provider offering one capability does not necessarily offer the others.

License

MIT