Image to Image
SAM 3
SAM 3 (Segment Anything Model 3) delivers foundation-level zero-shot visual segmentation and subject extraction. Seamlessly isolate objects, extract alpha-transparent PNG cutouts, or target specific items with natural language prompts.
API Endpoint
POST https://lumozai.com/api/v4/run
Use this endpoint to run the model with your parameters. Authentication is required using your API key.
Sample Code (PHP cURL)
<?php
// API endpoint URL
$apiUrl = 'https://lumozai.com/api/v4/run';
// Your API key
$apiKey = 'YOUR_API_KEY';
// Model parameters
$params = [
'model_id' => 'sam-3',
'image_url' => '',
'prompt' => '',
'apply_mask' => 'true',
'output_format' => 'png',
'return_multiple_masks' => 'false',
'max_masks' => 3,
'include_boxes' => 'false',
'include_scores' => 'false',
];
// Initialize cURL session
$curl = curl_init();
// Set cURL options
curl_setopt_array($curl, [
CURLOPT_URL => $apiUrl,
CURLOPT_RETURNTRANSFER => true,
CURLOPT_POST => true,
CURLOPT_POSTFIELDS => $params,
CURLOPT_HTTPHEADER => [
'Accept: application/json',
'Authorization: Bearer ' . $apiKey
]
]);
// Execute the request
$response = curl_exec($curl);
$httpCode = curl_getinfo($curl, CURLINFO_HTTP_CODE);
// Check for errors
if (curl_errno($curl)) {
echo 'cURL Error: ' . curl_error($curl);
} else {
// Process the response
$result = json_decode($response, true);
if ($httpCode == 200 && isset($result['success']) && $result['success']) {
// Handle successful response
echo "Model output: ";
print_r($result['data']['output']);
} else {
// Handle error response
echo "Error: " . ($result['message'] ?? 'Unknown error');
}
}
// Close cURL session
curl_close($curl);
Request Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
| model_id | string | Required | The ID of the model to run (sam-3) |
| api_key | string | Required | Your API key for authentication (sent via Bearer token in the Authorization header) |
| image_url | string | Required | Source image to segment or extract objects from. |
| prompt | string | Optional | Natural language prompt identifying which subject(s) to isolate. Leave blank to segment the primary subject. |
| apply_mask | string | Optional | Apply mask as a transparent cutout (alpha PNG) or export binary segmentation mask. |
| output_format | string | Optional | File format for the generated image or mask. |
| return_multiple_masks | string | Optional | Extract separate masks for each distinct detected object instance. |
| max_masks | number | Optional | Maximum number of masks returned when multiple masks is enabled. |
| include_boxes | string | Optional | Include bounding box coordinates in telemetry metadata. |
| include_scores | string | Optional | Include model segmentation confidence scores. |
Response Format
{
"success": true,
"data": {
"output": "https://lumozai.com/storage/outputs/image_12345.jpg"
}
}
Model Pricing
Pay only for what you run. Charges are deducted directly from your account balance with zero recurring commitments.
Rate per generation
$
0.005
/ generation
$5.00 per 1,000 generations
Cost Estimator
1,000 generations
Estimated Total
$5.00
Automatic instant refund if any run fails • Credits valid for 1 year