support >5h audio exports

src/tts.ts

@@ -1,5 +1,6 @@
-import { pipeline, RawAudio, TextToAudioPipeline } from "@huggingface/transformers";
+import { pipeline, TextToAudioPipeline } from "@huggingface/transformers";
 import { split } from "./splitter";
+import type { RawAudio } from "@huggingface/transformers";
 
 const MODEL_ID = "onnx-community/Supertonic-TTS-ONNX";
 const VOICES_URL = `https://huggingface.co/${MODEL_ID}/resolve/main/voices/`;
@@ -119,14 +120,35 @@ export async function* streamTTS(
 
 export function createAudioBlob(chunks: Float32Array[], sampling_rate: number): Blob {
   const totalLength = chunks.reduce((acc, chunk) => acc + chunk.length, 0);
-  const result = new Float32Array(totalLength);
-  let offset = 0;
-  for (const chunk of chunks) {
-    result.set(chunk, offset);
-    offset += chunk.length;
-  }
 
-  const audio = new RawAudio(result, sampling_rate);
-  const blob = audio.toBlob();
-  return blob;
+  // Create WAV header
+  const buffer = new ArrayBuffer(44);
+  const view = new DataView(buffer);
+
+  // RIFF chunk descriptor
+  writeString(view, 0, "RIFF");
+  view.setUint32(4, 36 + totalLength * 4, true); // ChunkSize
+  writeString(view, 8, "WAVE");
+
+  // fmt sub-chunk
+  writeString(view, 12, "fmt ");
+  view.setUint32(16, 16, true); // Subchunk1Size
+  view.setUint16(20, 3, true); // AudioFormat (3 = IEEE Float)
+  view.setUint16(22, 1, true); // NumChannels (Mono)
+  view.setUint32(24, sampling_rate, true); // SampleRate
+  view.setUint32(28, sampling_rate * 4, true); // ByteRate
+  view.setUint16(32, 4, true); // BlockAlign
+  view.setUint16(34, 32, true); // BitsPerSample
+
+  // data sub-chunk
+  writeString(view, 36, "data");
+  view.setUint32(40, totalLength * 4, true); // Subchunk2Size
+
+  return new Blob([buffer, ...chunks as any], { type: "audio/wav" });
+}
+
+function writeString(view: DataView, offset: number, string: string) {
+  for (let i = 0; i < string.length; i++) {
+    view.setUint8(offset + i, string.charCodeAt(i));
+  }
 }