Modification of the STT structure

apps/web/src/components/VoiceToTextButton.vue

@@ -137,6 +137,25 @@ function mergeChunks(parts: Float32Array[]) {
   return out;
 }
 
+/** Downsample to 16 kHz so STT does not depend on server ffmpeg. */
+function resampleTo16k(samples: Float32Array, srcRate: number) {
+  const dstRate = 16000;
+  if (!srcRate || srcRate === dstRate) return { samples, rate: srcRate || dstRate };
+  const ratio = srcRate / dstRate;
+  const newLen = Math.max(1, Math.floor(samples.length / ratio));
+  const out = new Float32Array(newLen);
+  for (let i = 0; i < newLen; i += 1) {
+    const srcIndex = i * ratio;
+    const i0 = Math.floor(srcIndex);
+    const i1 = Math.min(samples.length - 1, i0 + 1);
+    const t = srcIndex - i0;
+    const a = samples[i0] ?? 0;
+    const b = samples[i1] ?? a;
+    out[i] = a + (b - a) * t;
+  }
+  return { samples: out, rate: dstRate };
+}
+
 function encodeWav(samples: Float32Array, rate: number) {
   const bytes = samples.length * 2;
   const buffer = new ArrayBuffer(44 + bytes);
@@ -189,7 +208,8 @@ async function finish(keep: boolean) {
 
   setTranscribing(true);
   try {
-    const text = await upload(encodeWav(samples, rate));
+    const prepared = resampleTo16k(samples, rate);
+    const text = await upload(encodeWav(prepared.samples, prepared.rate));
     if (!text) {
       setError('متنی از صدا تشخیص داده نشد.');
       return;