diff --git a/ui/src/lib/components/AudioPlayer.svelte b/ui/src/lib/components/AudioPlayer.svelte
index 187318c..992ea7a 100644
--- a/ui/src/lib/components/AudioPlayer.svelte
+++ b/ui/src/lib/components/AudioPlayer.svelte
@@ -55,11 +55,11 @@
 		status = 'generating';
 		errorMsg = '';
 		try {
-			const res = await fetch(`http://localhost:8080/ui/audio/${slug}/${chapter}`, {
-				method: 'POST',
-				headers: { 'Content-Type': 'application/json' },
-				body: JSON.stringify({ voice, speed })
-			});
+		const res = await fetch(`/api/audio/${slug}/${chapter}`, {
+			method: 'POST',
+			headers: { 'Content-Type': 'application/json' },
+			body: JSON.stringify({ voice, speed })
+		});
 			if (!res.ok) throw new Error(`Generation failed: ${res.status}`);
 			// After generation, fetch the presigned URL
 			await loadAudio();
diff --git a/ui/src/routes/api/audio/[slug]/[n]/+server.ts b/ui/src/routes/api/audio/[slug]/[n]/+server.ts
new file mode 100644
index 0000000..c196bbe
--- /dev/null
+++ b/ui/src/routes/api/audio/[slug]/[n]/+server.ts
@@ -0,0 +1,92 @@
+import { error } from '@sveltejs/kit';
+import type { RequestHandler } from './$types';
+import { env } from '$env/dynamic/private';
+
+const SCRAPER_URL = env.SCRAPER_API_URL ?? 'http://localhost:8080';
+
+/**
+ * POST /api/audio/[slug]/[n]
+ * Proxies the audio generation request to the scraper's /api/audio endpoint.
+ * Keeps the scraper URL server-side — the browser never needs to know it.
+ *
+ * Body: { voice?: string, speed?: number }
+ * Response: { url: string, filename: string }
+ *   where `url` is a relative path to GET /api/audio/[slug]/[n]?voice=...&speed=...
+ */
+export const POST: RequestHandler = async ({ params, request }) => {
+	const { slug, n } = params;
+	const chapter = parseInt(n, 10);
+	if (!slug || !chapter || chapter < 1) {
+		error(400, 'Invalid slug or chapter number');
+	}
+
+	let body: { voice?: string; speed?: number } = {};
+	try {
+		body = await request.json();
+	} catch {
+		// empty body is fine — scraper will use defaults
+	}
+
+	const scraperRes = await fetch(`${SCRAPER_URL}/api/audio/${slug}/${chapter}`, {
+		method: 'POST',
+		headers: { 'Content-Type': 'application/json' },
+		body: JSON.stringify(body)
+	});
+
+	if (!scraperRes.ok) {
+		const text = await scraperRes.text().catch(() => '');
+		error(scraperRes.status as Parameters<typeof error>[0], text || 'Audio generation failed');
+	}
+
+	const data = (await scraperRes.json()) as { url: string; filename: string };
+
+	// Rewrite the proxy URL from the scraper's /ui/audio-proxy/... to our own
+	// /api/audio/[slug]/[n]?voice=...&speed=... so the browser never calls the scraper directly.
+	const voice = body.voice ?? '';
+	const speed = body.speed ?? 1.0;
+	const qs = new URLSearchParams();
+	if (voice) qs.set('voice', voice);
+	qs.set('speed', String(speed));
+
+	return new Response(
+		JSON.stringify({
+			url: `/api/audio/${slug}/${chapter}?${qs.toString()}`,
+			filename: data.filename
+		}),
+		{ headers: { 'Content-Type': 'application/json' } }
+	);
+};
+
+/**
+ * GET /api/audio/[slug]/[n]?voice=...&speed=...
+ * Proxies the audio stream from the scraper's /api/audio-proxy endpoint.
+ * This is the URL the browser's <audio> element uses as its src.
+ */
+export const GET: RequestHandler = async ({ params, url }) => {
+	const { slug, n } = params;
+	const chapter = parseInt(n, 10);
+	if (!slug || !chapter || chapter < 1) {
+		error(400, 'Invalid slug or chapter number');
+	}
+
+	const voice = url.searchParams.get('voice') ?? '';
+	const speed = url.searchParams.get('speed') ?? '1';
+	const qs = new URLSearchParams();
+	if (voice) qs.set('voice', voice);
+	qs.set('speed', speed);
+
+	const scraperRes = await fetch(`${SCRAPER_URL}/api/audio-proxy/${slug}/${chapter}?${qs.toString()}`);
+
+	if (!scraperRes.ok) {
+		error(scraperRes.status as Parameters<typeof error>[0], 'Audio not found');
+	}
+
+	// Stream the audio body through — preserve Content-Type and Content-Length.
+	const headers = new Headers();
+	headers.set('Content-Type', scraperRes.headers.get('Content-Type') ?? 'audio/mpeg');
+	headers.set('Cache-Control', 'public, max-age=3600');
+	const cl = scraperRes.headers.get('Content-Length');
+	if (cl) headers.set('Content-Length', cl);
+
+	return new Response(scraperRes.body, { headers });
+};