added caching to silence detection

Author: Matze99

src/sources/audio.spec.ts

@@ -70,6 +70,12 @@ describe('AudioSource', () => {
 		}]);
 	});
 
+	it('find silences correctly after caching', async () => {
+		const silences = await audioSource.silences({});
+		const cachedSilences = await audioSource.silences({threshold: 0, minDuration: 1e10, windowSize: 1e10});
+		expect(silences).toEqual(cachedSilences);
+	});
+
 	it('should decode an audio buffer correctly', async () => {
 		const buffer = await audioSource.decode(2, 44100, true);
 		expect(buffer.duration).toBe(5); // Mock duration

src/sources/audio.ts

@@ -18,6 +18,7 @@ const DEFAULT_SAMPLE_RATE = 3000;
 export class AudioSource<T extends Object = {}> extends Source<T> {
 	public readonly type: ClipType = 'audio';
 	private decoding = false;
+	private _silences?: { start: Timestamp; stop: Timestamp }[];
 
 	public transcript?: Transcript;
 	public audioBuffer?: AudioBuffer;
@@ -130,7 +131,7 @@ export class AudioSource<T extends Object = {}> extends Source<T> {
 	}
 
 	/**
-	 * Find silences in the audio clip
+	 * Find silences in the audio clip. Results are cached.
 	 * 
 	 * uses default sample rate of 3000
 	 * @param options - Silences options.
@@ -141,11 +142,14 @@ export class AudioSource<T extends Object = {}> extends Source<T> {
 		minDuration = 5,
 		windowSize = 50,
 	}: SilenceOptions): Promise<{ start: Timestamp; stop: Timestamp }[]> {
+		if (this._silences) return this._silences;
+
 		const audioBuffer = this.audioBuffer ?? (await this.decode(1, DEFAULT_SAMPLE_RATE, true));
 		const length = Math.floor(audioBuffer.length / windowSize);
 		const samples = await this.fastsampler({ length, logarithmic: false });
 
 		const silences = findSilences(samples, threshold, minDuration, this.duration.millis);
+		this._silences = silences;
 
 		return silences;
 	}

src/tracks/media/media.ts

@@ -9,7 +9,6 @@ import { Track } from '../track';
 
 import type { MediaClip } from '../../clips';
 import type { Timestamp } from '../../models';
-import { getSilenceArrayBuffer } from './media.utils';
 
 export class MediaTrack<Clip extends MediaClip> extends Track<MediaClip> {
 	public clips: Clip[] = [];
@@ -29,30 +28,23 @@ export class MediaTrack<Clip extends MediaClip> extends Track<MediaClip> {
      * @param minSilenceDuration Minimum duration in seconds for a silence period to be included
      * @returns Array of silence periods with start and stop times in seconds
      */
-    public async detectSilences(
+    public async removeSilences(
         subSample: number = 1000,
         silenceThreshold: number = -10, 
         minSilenceDuration: number = 0.1
-    ): Promise<{ start: number; stop: number }[]> {
-        const silences: { start: number; stop: number }[] = [];
-        const audioContext = new AudioContext();
+    ) {
 
         // Process each clip
         for (const clip of this.clips) {
             if (!clip.element) {
                 continue;
             }
 
-            // Get audio data for this clip
-            const arrayBuffer = await (await clip.source.getFile()).arrayBuffer();
-			const audioBuffer = await audioContext.decodeAudioData(arrayBuffer);
+            const silences = await clip.source.silences({});
+            if (silences.length === 0) continue;
 
-			silences.push(...getSilenceArrayBuffer(audioBuffer, subSample, minSilenceDuration, silenceThreshold, clip.start.seconds));
+            
 
         }
-
-        await audioContext.close();
-
-        return silences.sort((a, b) => a.start - b.start);
     }
 }