Hamza/nabeh
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Implement 2-step audio-to-audio process

Browse Source
This commit is contained in:
Hamza-Ayed
2026-05-22 15:29:30 +03:00
parent ffd8c6f2a5
commit f1d57e2763
2 changed files with 22 additions and 133 deletions
Download Patch File Download Diff File Expand all files Collapse all files

68
backend/public/test_audio.php
View File

@@ -135,71 +135,19 @@ if ($successfulModel && $audioResponse && !empty($audioResponse['audio'])) {
echo "ℹ️ [Gemini] Audio MimeType: " . $audioResponse['mimeType'] . "\n";
echo "ℹ️ [Gemini] Audio Size: " . strlen($audioResponse['audio']) . " base64 chars\n";
// 3. Test Audio-to-Audio conversion using the successful model
echo "\n--- Testing Audio-to-Audio (Speech-to-Speech) ---\n";
// 3. Test Audio-to-Audio conversion using the Service
echo "\n--- Testing Audio-to-Audio (Speech-to-Speech) via GeminiService ---\n";
$startTime = microtime(true);
// We temporarily override the model inside GeminiService for this test, but since GeminiService is not updated yet,
// we'll run a direct curl call for testing audio-to-audio:
$url = 'https://generativelanguage.googleapis.com/v1beta/models/' . $successfulModel . ':generateContent?key=' . $apiKey;
$payload2 = json_encode([
'contents' => [
[
'role' => 'user',
'parts' => [
[
'inlineData' => [
'mimeType' => $audioResponse['mimeType'],
'data' => $audioResponse['audio']
]
],
[
'text' => "استمع إلى التسجيل الصوتي المرفق وأجب عليه مباشرة بصوتك بناءً على الإرشادات المحددة."
]
]
]
],
'systemInstruction' => [
'parts' => [['text' => $systemPrompt]]
],
'generationConfig' => [
'responseModalities' => ['AUDIO'],
'speechConfig' => [
'voiceConfig' => [
'prebuiltVoiceConfig' => [
'voiceName' => 'Puck'
]
]
]
]
]);
$ch = curl_init($url);
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload2);
curl_setopt($ch, CURLOPT_HTTPHEADER, ['Content-Type: application/json']);
curl_setopt($ch, CURLOPT_TIMEOUT, 30);
$response2 = curl_exec($ch);
$httpCode2 = curl_getinfo($ch, CURLINFO_HTTP_CODE);
curl_close($ch);
$audioResponse2 = GeminiService::generateAudioResponseFromAudio($apiKey, $systemPrompt, $audioResponse['audio'], $audioResponse['mimeType'], 'Puck');
$elapsedTime2 = round(microtime(true) - $startTime, 2);
if ($httpCode2 === 200) {
$data2 = json_decode($response2, true);
$part2 = $data2['candidates'][0]['content']['parts'][0] ?? null;
if ($part2 && isset($part2['inlineData'])) {
echo "✅ [Gemini] Successfully generated Audio-to-Audio response in {$elapsedTime2} seconds!\n";
echo "ℹ️ [Gemini] Audio MimeType: " . ($part2['inlineData']['mimeType'] ?? 'audio/mp4') . "\n";
echo "ℹ️ [Gemini] Audio Size: " . strlen($part2['inlineData']['data']) . " base64 chars\n";
} else {
echo "❌ [Gemini] Audio-to-Audio response did not contain inlineData.\n";
}
if ($audioResponse2 && !empty($audioResponse2['audio'])) {
echo "✅ [Gemini] Successfully generated Audio-to-Audio response in {$elapsedTime2} seconds!\n";
echo "ℹ️ [Gemini] Audio MimeType: " . $audioResponse2['mimeType'] . "\n";
echo "ℹ️ [Gemini] Audio Size: " . strlen($audioResponse2['audio']) . " base64 chars\n";
} else {
$errorData2 = json_decode($response2, true);
$errMsg2 = $errorData2['error']['message'] ?? 'Unknown error';
echo "❌ [Gemini] Audio-to-Audio generation failed. HTTP {$httpCode2}: {$errMsg2}\n";
echo "❌ [Gemini] Audio-to-Audio generation failed.\n";
}
} else {
echo "\n❌ [Gemini] All model trials for audio response generation failed.\n";