Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/phpaisdk/openai/llms.txt

Use this file to discover all available pages before exploring further.

The aisdk/openai package supports multimodal inputs — text, images (URL or base64 data URI), audio clips (base64-encoded), and file attachments — by mapping PHP Content objects into the OpenAI Chat Completions messages array format. Each content type is converted transparently by ChatMessageConverter, so you work with the portable SDK API and the provider handles the wire format.

​
Image Input

Attach one or more images to a user message by mixing Content::text() and Content::image() inside Message::user([...]). The SDK converts each image to the OpenAI image_url content part format. 
use AiSdk\Content;
use AiSdk\Generate;
use AiSdk\Message;
use AiSdk\OpenAI;

$result = Generate::text()
    ->messages([
        Message::user([
            Content::text('What is this?'),
            Content::image('https://example.com/photo.png'),
        ]),
    ])
    ->model(OpenAI::model('gpt-4o'))
    ->run();

echo $result->text;
The image is sent to the API as: 
{
  "type": "image_url",
  "image_url": { "url": "https://example.com/photo.png" }
}
Images can be provided as either a remote URL or a base64 data URI. When you pass a base64-encoded image, the SDK constructs the data:<mime>;base64,<data> URI and places it in the url field — OpenAI accepts both formats interchangeably. Compatible models (from models.json): gpt-4o, gpt-4o*, gpt-4.1*, o*, gpt-5*.

​
Audio Input

Send audio content alongside a text prompt using Content::audio(). The SDK converts the clip to OpenAI’s input_audio content part, selecting the correct format name from the MIME type automatically. 
use AiSdk\Content;
use AiSdk\Generate;
use AiSdk\Message;
use AiSdk\OpenAI;

$audioBase64 = base64_encode(file_get_contents('/path/to/clip.mp3'));

$result = Generate::text()
    ->messages([
        Message::user([
            Content::text('Transcribe and summarise this recording.'),
            Content::audio($audioBase64, mimeType: 'audio/mpeg'),
        ]),
    ])
    ->model(OpenAI::model('gpt-4o-audio-preview'))
    ->run();

echo $result->text;

​
Supported Audio Formats

The SDK maps MIME types to the format names OpenAI expects in the input_audio.format field:
MIME TypeOpenAI Format
audio/mpegmp3
audio/mp3mp3
audio/wavwav
audio/x-wavwav
audio/oggogg
(any other)wav (default fallback)
Compatible models: patterns matching gpt-*-audio* (e.g. gpt-4o-audio-preview).

​
File Input

Attach PDF documents or other files to a message using Content::file(). The SDK converts the attachment to OpenAI’s file content part with a file_data field containing the base64 data URI. 
use AiSdk\Content;
use AiSdk\Generate;
use AiSdk\InputEncoding;
use AiSdk\Message;
use AiSdk\OpenAI;

$result = Generate::text()
    ->messages([
        Message::user([
            Content::text('Summarise the key findings in this report.'),
            Content::file(
                'JVBERi0=',
                mimeType: 'application/pdf',
                filename: 'report.pdf',
                encoding: InputEncoding::Base64,
            ),
        ]),
    ])
    ->model(OpenAI::model('gpt-4.1'))
    ->run();

echo $result->text;
The file is sent to the API as: 
{
  "type": "file",
  "file": {
    "filename": "report.pdf",
    "file_data": "data:application/pdf;base64,JVBERi0="
  }
}
Compatible models: gpt-4.1*, gpt-5*.
Passing file input to a model that does not have the FileInput capability — for example gpt-4o — throws a CapabilityNotSupportedException before making any HTTP request. Always verify capability support or use a compatible model.

​
Capability Checking

Before sending a multimodal message, you can interrogate the model handle to confirm it supports the content types you intend to send. This avoids an unnecessary round-trip when the capability is absent. 
use AiSdk\Capability;
use AiSdk\OpenAI;

$model = OpenAI::model('gpt-4.1');

if ($model->supports(Capability::ImageInput)) {
    // Safe to attach Content::image(...)
}

if ($model->supports(Capability::AudioInput)) {
    // Safe to attach Content::audio(...)
}

if ($model->supports(Capability::FileInput)) {
    // Safe to attach Content::file(...)
}

​
Combining Multiple Content Types

Multiple content types can coexist in the same user message. Here is an example that combines descriptive text with an image URL in a single turn: 
use AiSdk\Content;
use AiSdk\Generate;
use AiSdk\Message;
use AiSdk\OpenAI;

$result = Generate::text()
    ->messages([
        Message::user([
            Content::text('Describe what you see in this diagram and explain it simply.'),
            Content::image('https://example.com/architecture-diagram.png'),
        ]),
    ])
    ->model(OpenAI::model('gpt-4o'))
    ->run();

echo $result->text;
Use Message::user([...]) with an array of Content objects whenever you need mixed content in one turn. For a plain text-only message you can pass a string directly: Message::user('Hello').

Build docs developers (and LLMs) love

Get started for freeTalk to us