diff --git a/ext/mixed/js/audio.js b/ext/mixed/js/audio.js
index b5a025be..6a338cca 100644
--- a/ext/mixed/js/audio.js
+++ b/ext/mixed/js/audio.js
@@ -83,6 +83,124 @@ class TextToSpeechAudio {
     }
 }
 
+class AudioSystem {
+    constructor() {
+        this._cache = new Map();
+        this._cacheSizeMaximum = 32;
+
+        if (typeof speechSynthesis !== 'undefined') {
+            // speechSynthesis.getVoices() will not be populated unless some API call is made.
+            speechSynthesis.addEventListener('voiceschanged', this._onVoicesChanged.bind(this));
+        }
+    }
+
+    async getExpressionAudio(expression, sources, optionsContext, details) {
+        const key = `${expression.expression}:${expression.reading}`;
+        const cacheValue = this._cache.get(expression);
+        if (typeof cacheValue !== 'undefined') {
+            const {audio, uri, source} = cacheValue;
+            return {audio, uri, source};
+        }
+
+        for (const source of sources) {
+            const uri = await apiAudioGetUrl(expression, source, optionsContext);
+            if (uri === null) { continue; }
+
+            try {
+                const audio = await this._createAudio(uri, details);
+                this._cacheCheck();
+                this._cache.set(key, {audio, uri, source});
+                return {audio, uri, source};
+            } catch (e) {
+                // NOP
+            }
+        }
+
+        throw new Error('Could not create audio');
+    }
+
+    createTextToSpeechAudio({text, voiceUri}) {
+        const voice = this._getTextToSpeechVoiceFromVoiceUri(voiceUri);
+        if (voice === null) {
+            throw new Error('Invalid text-to-speech voice');
+        }
+        return new TextToSpeechAudio(text, voice);
+    }
+
+    _onVoicesChanged() {
+        // NOP
+    }
+
+    async _createAudio(uri, details) {
+        const ttsParameters = this._getTextToSpeechParameters(uri);
+        if (ttsParameters !== null) {
+            if (typeof details === 'object' && details !== null) {
+                if (details.tts === false) {
+                    throw new Error('Text-to-speech not permitted');
+                }
+            }
+            return this.createTextToSpeechAudio(ttsParameters);
+        }
+
+        return await this._createAudioFromUrl(uri);
+    }
+
+    _createAudioFromUrl(url) {
+        return new Promise((resolve, reject) => {
+            const audio = new Audio(url);
+            audio.addEventListener('loadeddata', () => {
+                const duration = audio.duration;
+                if (duration === 5.694694 || duration === 5.720718) {
+                    // Hardcoded values for invalid audio
+                    reject(new Error('Could not retrieve audio'));
+                } else {
+                    resolve(audio);
+                }
+            });
+            audio.addEventListener('error', () => reject(audio.error));
+        });
+    }
+
+    _getTextToSpeechVoiceFromVoiceUri(voiceUri) {
+        try {
+            for (const voice of speechSynthesis.getVoices()) {
+                if (voice.voiceURI === voiceUri) {
+                    return voice;
+                }
+            }
+        } catch (e) {
+            // NOP
+        }
+        return null;
+    }
+
+    _getTextToSpeechParameters(uri) {
+        const m = /^tts:[^#?]*\?([^#]*)/.exec(uri);
+        if (m === null) { return null; }
+
+        const searchParameters = new URLSearchParams(m[1]);
+        const text = searchParameters.get('text');
+        const voiceUri = searchParameters.get('voice');
+        return (text !== null && voiceUri !== null ? {text, voiceUri} : null);
+    }
+
+    _cacheCheck() {
+        const removeCount = this._cache.size - this._cacheSizeMaximum;
+        if (removeCount <= 0) { return; }
+
+        const removeKeys = [];
+        for (const key of this._cache.keys()) {
+            removeKeys.push(key);
+            if (removeKeys.length >= removeCount) { break; }
+        }
+
+        for (const key of removeKeys) {
+            this._cache.delete(key);
+        }
+    }
+}
+
+
 function audioGetFromUrl(url, willDownload) {
     const tts = TextToSpeechAudio.createFromUri(url);
     if (tts !== null) {