Fix ASR

src/audio.ts

@@ -1,6 +1,8 @@
 import {spawn} from 'node:child_process';
-import {pipeline, read_audio} from '@xenova/transformers';
+import {pipeline} from '@xenova/transformers';
+import * as fs from 'node:fs';
 import {AbortablePromise, Ai} from './ai.ts';
+import * as wavefile from 'wavefile';
 
 export class Audio {
 	private whisperPipeline: any;
@@ -96,8 +98,12 @@ print(json.dumps(segments))
 
 				// Transcript
 				if(aborted) return resolve(null);
-				const audio = await read_audio(path, 16000);
-				const transcriptResult = await this.whisperPipeline(audio, {return_timestamps: speaker ? 'word' : false, chunk_length_s: 30,});
+				const audioData = fs.readFileSync(path);
+				const wav = new wavefile.WaveFile(audioData);
+				wav.toBitDepth('32f');
+				wav.toSampleRate(16000);
+				const buffer = wav.getSamples();
+				const transcriptResult = await this.whisperPipeline(buffer, {return_timestamps: speaker ? 'word' : false, chunk_length_s: 30,});
 				if(!speaker) return resolve(transcriptResult.text?.trim() || null);
 
 				// Speaker Diarization