Since you already have a string (the curl result body) you can use PHPs XPath or DOMDocument classes to extract some attribute values.
<?php
$html = '<div class="mdl-detail-song">
<div class="title">
<span href="javascript:void(0);" class="txt>title</span>
<div class="txt-singer"></div>
<input type="hidden" id="audio_file_path" value="example.mp3">';
libxml_use_internal_errors(true);
$dom = new DOMDocument();
$dom->omitXmlDeclaration = true;
$dom->preserveWhiteSpace = false;
$dom->validateOnParse = false;
$dom->strictErrorChecking = false;
$dom->formatOutput = false;
$dom->loadHTML('<?xml encoding="utf-8" ?>'.$html);
libxml_clear_errors();
libxml_use_internal_errors(false);
// use DOMDocument
$input = $dom->getElementsByTagName('input')->item(0); // first input in HTML
echo $input->getAttribute('id'); // audio_file_path
// use XPath
$xpath = new DOMXPath($dom);
$input = $xpath->query('(//input[@id]/@id)[1]'); // attribute id of first input which has an id attribute
echo $input->item(0)->nodeValue; // audio_file_path
Note that the HTML your trying to parse isn't complete or simply broken (which is why I added some checks you wouldn't need when reading proper HTML). This way you can read and parse strict and non-strict or html5ish html without problems.