diff --git a/realtime/index.html b/realtime/index.html index df02b6e..fea3d8a 100644 --- a/realtime/index.html +++ b/realtime/index.html @@ -36,6 +36,10 @@ 🤖 Voice Agent Demo Lets you chat with a simple voice agent. + + ⏩ Speed Control Demo + Adjust speech speed while talking to the agent. + 📓 Voice to Text Agent Demo Lets you chat with an agent that listens to voice and replies with text. diff --git a/realtime/speed/index.html b/realtime/speed/index.html new file mode 100644 index 0000000..aca5897 --- /dev/null +++ b/realtime/speed/index.html @@ -0,0 +1,64 @@ + + + Realtime Speed Demo + + + + + +
+
+ Realtime Demos + ⏩ Speed Control +
+
+ This demo shows how to use the OpenAI Realtime API to create an interactive voice agent. + Use the slider to control the speaking speed. +
+
+ + +
+
+ + +
+
+ + +
+
+ + +
+
+ + +
+
+
+ + +
+
+ +
+
+

GitHub

+
+ + + + diff --git a/realtime/speed/main.js b/realtime/speed/main.js new file mode 100644 index 0000000..9a1c04b --- /dev/null +++ b/realtime/speed/main.js @@ -0,0 +1,103 @@ +const APP_PREFIX = "realtime/speed/"; +const $ = document.querySelector.bind(document); +const apiKeyEl = $("#openai-api-key"); +const modelEl = $("#model"); +const voiceEl = $("#voice"); +const speedEl = $("#speed"); +const instructionsEl = $("#instructions"); +const startMicrophoneEl = $("#start-microphone"); +const stopEl = $("#stop"); +const statusEl = $("#status"); +const prefs = [apiKeyEl, modelEl, voiceEl, instructionsEl]; + +let session = null; + +function initState() { + prefs.forEach(p => { + const fqid = p.id != "openai-api-key" ? APP_PREFIX + p.id : p.id; + const v = localStorage.getItem(fqid); + if (v) { + p.value = v; + } + p.addEventListener("change", () => { + localStorage.setItem(fqid, p.value); + }); + }); + const speedVal = localStorage.getItem(APP_PREFIX + "speed"); + if (speedVal) { + speedEl.value = speedVal; + } + speedEl.addEventListener("change", handleSpeedChange); + updateState(false); +} + +function updateState(started) { + statusEl.textContent = ""; + prefs.forEach(p => p.disabled = started); + startMicrophoneEl.disabled = started; + stopEl.disabled = !started; +} + +async function startMicrophone() { + if (!apiKeyEl.value) { + window.alert("Please enter your OpenAI API Key. You can obtain one from https://platform.openai.com/settings/organization/api-keys"); + return; + } + const stream = await navigator.mediaDevices.getUserMedia({ audio: true }); + start(stream); +} + +async function start(stream) { + updateState(true); + session = new Session(apiKeyEl.value); + session.onconnectionstatechange = state => statusEl.textContent = state; + session.ontrack = e => handleTrack(e); + session.onopen = e => handleOpen(); + session.onmessage = e => handleMessage(e); + session.onerror = e => handleError(e); + const sessionConfig = { + model: modelEl.value, + voice: voiceEl.value, + speed: parseFloat(speedEl.value), + instructions: instructionsEl.value || undefined + } + await session.start(stream, sessionConfig); +} + +function stop() { + updateState(false); + session.stop(); + session = null; +} + +function handleTrack(e) { + const audio = new Audio(); + audio.srcObject = e.streams[0]; + audio.play(); +} + +function handleSpeedChange() { + localStorage.setItem(APP_PREFIX + "speed", speedEl.value); + if (session) { + const updateMessage = { type: "session.update", session: { speed: parseFloat(speedEl.value) } }; + session.sendMessage(updateMessage); + const createResponse = { type: "response.create" }; + session.sendMessage(createResponse); + } +} + +function handleOpen() { + const message = { type: "response.create" }; + session.sendMessage(message); +} + +function handleMessage(message) { + console.log("message", message); +} + +function handleError(e) { + console.error(e); + stop(); +} + +initState();