echogarden 2.10.2 → 3.0.1

package/src/synthesis/EspeakTTS.ts

@@ -4,25 +4,34 @@ import { Logger } from '../utilities/Logger.js'
 import { RawAudio, getEmptyRawAudio } from '../audio/AudioUtilities.js'
 import { getNormalizedFragmentsForSpeech, simplifyPunctuationCharacters } from '../nlp/TextNormalizer.js'
 import { ipaPhoneToKirshenbaum } from '../nlp/PhoneConversion.js'
-import { splitToWords, wordCharacterRegExp } from '../nlp/Segmentation.js'
+import { includesPunctuation, splitToWords, wordCharacterRegExp } from '../nlp/Segmentation.js'
 import { Lexicon, tryGetFirstLexiconSubstitution } from '../nlp/Lexicon.js'
 import { phonemizeSentence } from '../nlp/EspeakPhonemizer.js'
 import { Timeline, TimelineEntry } from '../utilities/Timeline.js'
 import { extendDeep } from '../utilities/ObjectUtilities.js'
 import { escapeHtml } from '../encodings/HtmlEscape.js'
+import * as TextSegmentation from '@echogarden/text-segmentation'
+
+import { getGlobalOption, OperationCallbacks, SynthesisCallbacks } from '../api/API.js'
+import { loadPackage } from '../utilities/PackageManager.js'
 
 import { wrapEmscriptenModuleHeap } from 'wasm-heap-manager'
+import { pathToFileUrl, resolvePath } from '../utilities/PathUtilities.js'
 
 const log = logToStderr
 
 let espeakInstance: any
 let espeakModule: any
 
-export async function preprocessAndSynthesize(text: string, language: string, espeakOptions: EspeakOptions, lexicons: Lexicon[] = []) {
-	const logger = new Logger()
+export async function preprocessAndSynthesize(text: string, language: string, espeakOptions: EspeakOptions, lexicons: Lexicon[] = [], callbacks: SynthesisCallbacks) {
+	const logger = new Logger(callbacks.logLevel)
 
 	espeakOptions = extendDeep(defaultEspeakOptions, espeakOptions)
 
+	await logger.startAsync('Load espeak instance')
+
+	await ensureEspeakInstanceInitialized(callbacks)
+
 	await logger.startAsync('Tokenize and analyze text')
 
 	let lowerCaseLanguageCode = language.toLowerCase()
@@ -42,7 +51,7 @@ export async function preprocessAndSynthesize(text: string, language: string, es
 
 	// Merge repeating non-words to a single word to work around eSpeak bug
 	{
-		const wordsWithMerges: string[] = []
+		const mergedWords: string[] = []
 
 		for (let i = 0; i < words.length; i++) {
 			const currentWord = words[i]
@@ -54,13 +63,22 @@ export async function preprocessAndSynthesize(text: string, language: string, es
 				!['[', ']'].includes(currentWord) && // Work around eSpeak-NG marker bug with repeating squared brackets
 				!wordCharacterRegExp.test(currentWord)) {
 
-				wordsWithMerges[wordsWithMerges.length - 1] += currentWord
+				mergedWords[mergedWords.length - 1] += currentWord
 			} else {
-				wordsWithMerges.push(currentWord)
+				mergedWords.push(currentWord)
+			}
+		}
+
+		for (let i = 0; i < mergedWords.length; i++) {
+			const mergedWord = mergedWords[i]
+
+			// Replace vertical bar only words with empty words
+			if (/^[\|│]+$/.test(mergedWord)) {
+				mergedWords[i] = ''
 			}
 		}
 
-		words = wordsWithMerges
+		words = mergedWords
 	}
 
 	// Remove words containing only whitespace
@@ -97,8 +115,8 @@ export async function preprocessAndSynthesize(text: string, language: string, es
 		}
 
 		phonemizedFragmentsSubstitutions.set(fragmentIndex, substitutionPhonemes)
-		const referenceIPA = (await textToPhonemes(fragment, espeakOptions.voice, true)).replaceAll('_', ' ')
-		const referenceKirshenbaum = (await textToPhonemes(fragment, espeakOptions.voice, false)).replaceAll('_', '')
+		const referenceIPA = textToPhonemes(fragment, espeakOptions.voice, true).replaceAll('_', ' ')
+		const referenceKirshenbaum = textToPhonemes(fragment, espeakOptions.voice, false).replaceAll('_', '')
 
 		const kirshenbaumPhonemes = substitutionPhonemes.map(phone => ipaPhoneToKirshenbaum(phone)).join('')
 
@@ -114,42 +132,95 @@ export async function preprocessAndSynthesize(text: string, language: string, es
 
 	logger.start('Synthesize preprocessed fragments with eSpeak')
 
-	const { rawAudio: referenceSynthesizedAudio, timeline: referenceTimeline } = await synthesizeFragments(preprocessedFragments, espeakOptions)
+	const { rawAudio: referenceSynthesizedAudio, timeline: referenceTimeline } = await synthesizeFragments(
+		preprocessedFragments,
+		espeakOptions,
+		{ ...callbacks, logLevel: 'warning' }
+	)
 
 	await logger.startAsync('Build phonemized tokens')
 
+	// Build modified clause timeline using word segmentation library instead
+	// of eSpeak's native clauses
+	const clauseTimeline: Timeline = []
+
+	{
+		const fragmentWordSequence = new TextSegmentation.WordSequence()
+
+		for (let fragment of fragments) {
+			fragmentWordSequence.addWord(fragment, 0, !wordCharacterRegExp.test(fragment))
+		}
+
+		const wordEntries = referenceTimeline.flatMap(phraseEntry => phraseEntry.timeline!)
+		const segmentedWords =
+			await TextSegmentation.segmentWordSequence(fragmentWordSequence, {
+				requireSpaceAfterColonsOrSemicolons: false
+			})
+
+		let phraseStartWordIndex = 0
+
+		for (const sentence of segmentedWords.sentences) {
+			for (const phrase of sentence.phrases) {
+				const phraseEndWordIndex = phraseStartWordIndex + phrase.words.length
+
+				const phraseEntryTimeline: Timeline = []
+
+				for (let wordIndex = phraseStartWordIndex; wordIndex < phraseEndWordIndex; wordIndex++) {
+					phraseEntryTimeline.push(wordEntries[wordIndex])
+				}
+
+				const clauseEntry: TimelineEntry = {
+					type: 'clause',
+					text: phraseEntryTimeline.map(wordEntry => wordEntry.text).join(' '),
+					startTime: phraseEntryTimeline[0].startTime,
+					endTime: phraseEntryTimeline[phraseEntryTimeline.length - 1].endTime,
+					timeline: phraseEntryTimeline,
+				}
+
+				clauseTimeline.push(clauseEntry)
+
+				phraseStartWordIndex = phraseEndWordIndex
+			}
+		}
+	}
+
 	const phonemizedSentence: string[][][] = []
 
-	let wordIndex = 0
-	for (const phraseEntry of referenceTimeline) {
-		const phrase: string[][] = []
+	{
+		let wordIndex = 0
 
-		for (const wordEntry of phraseEntry.timeline!) {
-			wordEntry.text = fragments[wordIndex]
+		for (const phraseEntry of clauseTimeline) {
+			const phrase: string[][] = []
 
-			if (phonemizedFragmentsSubstitutions.has(wordIndex)) {
-				phrase.push(phonemizedFragmentsSubstitutions.get(wordIndex)!)
-			} else {
-				for (const tokenEntry of wordEntry.timeline!) {
-					const tokenPhonemes: string[] = []
+			for (const wordEntry of phraseEntry.timeline!) {
+				wordEntry.text = fragments[wordIndex]
+
+				if (phonemizedFragmentsSubstitutions.has(wordIndex)) {
+					phrase.push(phonemizedFragmentsSubstitutions.get(wordIndex)!)
+				} else {
+					for (const tokenEntry of wordEntry.timeline!) {
+						const tokenPhonemes: string[] = []
 
-					for (const phoneme of tokenEntry.timeline!) {
-						if (phoneme.text) {
-							tokenPhonemes.push(phoneme.text)
+						for (const phoneme of tokenEntry.timeline!) {
+							if (phoneme.text) {
+								tokenPhonemes.push(phoneme.text)
+							}
 						}
-					}
 
-					if (tokenPhonemes.length > 0) {
-						phrase.push(tokenPhonemes)
+						if (tokenPhonemes.length > 0) {
+							phrase.push(tokenPhonemes)
+						}
 					}
 				}
+
+				wordIndex += 1
 			}
 
-			wordIndex += 1
-		}
+			phraseEntry.text = phraseEntry.timeline!.map(wordEntry => wordEntry.text).join(' ')
 
-		if (phrase.length > 0) {
-			phonemizedSentence.push(phrase)
+			if (phrase.length > 0) {
+				phonemizedSentence.push(phrase)
+			}
 		}
 	}
 
@@ -157,15 +228,15 @@ export async function preprocessAndSynthesize(text: string, language: string, es
 
 	logger.end()
 
-	return { referenceSynthesizedAudio, referenceTimeline, fragments, preprocessedFragments, phonemizedFragmentsSubstitutions, phonemizedSentence }
+	return { referenceSynthesizedAudio, referenceTimeline: clauseTimeline, fragments, preprocessedFragments, phonemizedFragmentsSubstitutions, phonemizedSentence }
 }
 
-export async function synthesizeFragments(fragments: string[], espeakOptions: EspeakOptions) {
+export async function synthesizeFragments(fragments: string[], espeakOptions: EspeakOptions, callbacks: SynthesisCallbacks) {
 	espeakOptions = extendDeep(defaultEspeakOptions, espeakOptions)
 
 	const voice = espeakOptions.voice
 
-	const sampleRate = await getSampleRate()
+	const sampleRate = getSampleRate()
 
 	if (fragments.length === 0) {
 		return {
@@ -210,7 +281,7 @@ export async function synthesizeFragments(fragments: string[], espeakOptions: Es
 		}
 	}
 
-	const { rawAudio, events } = await synthesize(textWithMarkers, { ...espeakOptions, ssml: true })
+	const { rawAudio, events } = await synthesize(textWithMarkers, { ...espeakOptions, ssml: true }, callbacks)
 
 	// Add first marker if missing
 	if (fragments.length > 0) {
@@ -352,7 +423,7 @@ export async function synthesizeFragments(fragments: string[], espeakOptions: Es
 			continue
 		}
 
-		const wordReferencePhonemes = (await textToPhonemes(wordEntry.text, espeakOptions.voice, true)).split('_')
+		const wordReferencePhonemes = textToPhonemes(wordEntry.text, espeakOptions.voice, true).split('_')
 
 		const wordReferenceIPA = wordReferencePhonemes.join(' ')
 
@@ -449,8 +520,8 @@ export async function synthesizeFragments(fragments: string[], espeakOptions: Es
 	return { rawAudio, timeline: clauseTimeline, events }
 }
 
-export async function synthesize(text: string, espeakOptions: EspeakOptions) {
-	const logger = new Logger()
+export async function synthesize(text: string, espeakOptions: EspeakOptions, callbacks: SynthesisCallbacks) {
+	const logger = new Logger(callbacks.logLevel)
 
 	espeakOptions = extendDeep(defaultEspeakOptions, espeakOptions)
 
@@ -460,7 +531,7 @@ export async function synthesize(text: string, espeakOptions: EspeakOptions) {
 		text = escapeHtml(text)
 	}
 
-	const { instance } = await getEspeakInstance()
+	const { instance } = await getEspeakInstance(callbacks)
 
 	const sampleChunks: Float32Array[] = []
 	const allEvents: EspeakEvent[] = []
@@ -468,14 +539,14 @@ export async function synthesize(text: string, espeakOptions: EspeakOptions) {
 	logger.start('Synthesize with eSpeak')
 
 	if (espeakOptions.useKlatt) {
-		await setVoice(`${espeakOptions.voice}+klatt6`)
+		setVoice(`${espeakOptions.voice}+klatt6`)
 	} else {
-		await setVoice(espeakOptions.voice)
+		setVoice(espeakOptions.voice)
 	}
 
-	await setRate(espeakOptions.rate)
-	await setPitch(espeakOptions.pitch)
-	await setPitchRange(espeakOptions.pitchRange)
+	setRate(espeakOptions.rate)
+	setPitch(espeakOptions.pitch)
+	setPitchRange(espeakOptions.pitchRange)
 
 	instance.synthesize(text, (samples: Int16Array, events: EspeakEvent[]) => {
 		if (samples && samples.length > 0) {
@@ -501,22 +572,24 @@ export async function synthesize(text: string, espeakOptions: EspeakOptions) {
 	return { rawAudio, events: allEvents }
 }
 
-export async function textToIPA(text: string, voice: string) {
-	await setVoice(voice)
+export function textToIPA(text: string, voice: string) {
+	errorIfEspeakInstanceNotLoaded()
+
+	setVoice(voice)
 
-	const { instance } = await getEspeakInstance()
-	const ipa: string = (instance.synthesize_ipa(text).ipa as string).trim()
+	const ipa: string = (espeakInstance.synthesize_ipa(text).ipa as string).trim()
 
 	return ipa
 }
 
-export async function textToPhonemes(text: string, voice: string, useIPA = true) {
-	await setVoice(voice)
+export function textToPhonemes(text: string, voice: string, useIPA = true) {
+	errorIfEspeakInstanceNotLoaded()
 
-	const { instance, module } = await getEspeakInstance()
-	const textPtr = instance.convert_to_phonemes(text, useIPA)
+	setVoice(voice)
 
-	const wasmHeap = wrapEmscriptenModuleHeap(module)
+	const textPtr = espeakInstance.convert_to_phonemes(text, useIPA)
+
+	const wasmHeap = wrapEmscriptenModuleHeap(espeakModule)
 
 	const resultRef = wasmHeap.wrapNullTerminatedUtf8String(textPtr.ptr)
 	const result = resultRef.value
@@ -527,46 +600,46 @@ export async function textToPhonemes(text: string, voice: string, useIPA = true)
 }
 
 let lastVoiceId: string | undefined
-export async function setVoice(voiceId: string) {
-	const { instance } = await getEspeakInstance()
+export function setVoice(voiceId: string) {
+	errorIfEspeakInstanceNotLoaded()
 
 	if (voiceId !== lastVoiceId) {
-		instance.set_voice(voiceId)
+		espeakInstance.set_voice(voiceId)
 
 		lastVoiceId = voiceId
 	}
 }
 
-export async function setVolume(volume: number) {
-	const { instance } = await getEspeakInstance()
+export function setVolume(volume: number) {
+	errorIfEspeakInstanceNotLoaded()
 
-	return instance.setVolume(volume)
+	return espeakInstance.setVolume(volume)
 }
 
 export async function setRate(rate: number) {
-	const { instance } = await getEspeakInstance()
+	errorIfEspeakInstanceNotLoaded()
 
-	return instance.set_rate(rate)
+	return espeakInstance.set_rate(rate)
 }
 
 export async function setPitch(pitch: number) {
-	const { instance } = await getEspeakInstance()
+	errorIfEspeakInstanceNotLoaded()
 
-	return instance.set_pitch(pitch)
+	return espeakInstance.set_pitch(pitch)
 }
 
 export async function setPitchRange(pitchRange: number) {
-	const { instance } = await getEspeakInstance()
+	errorIfEspeakInstanceNotLoaded()
 
-	return instance.set_range(pitchRange)
+	return espeakInstance.set_range(pitchRange)
 }
 
-export async function getSampleRate(): Promise<22050> {
+export function getSampleRate(): 22050 {
 	return 22050
 }
 
-export async function listVoices() {
-	const { instance } = await getEspeakInstance()
+export async function listVoices(callbacks: OperationCallbacks) {
+	const { instance } = await getEspeakInstance(callbacks)
 
 	const voiceList: {
 		identifier: string,
@@ -580,9 +653,19 @@ export async function listVoices() {
 	return voiceList
 }
 
-async function getEspeakInstance() {
+function errorIfEspeakInstanceNotLoaded() {
 	if (!espeakInstance) {
-		const { default: EspeakInitializer } = await import('@echogarden/espeak-ng-emscripten')
+		throw new Error(`Espeak instance has not been initialized.`)
+	}
+}
+
+async function getEspeakInstance(callbacks: OperationCallbacks) {
+	if (!espeakInstance) {
+		const packagePath = await loadPackage('espeak-ng-emscripten', callbacks)
+		const normalizedModulePath = resolvePath(packagePath, 'espeak-ng.js')
+		const modulePathAsFileUrl = pathToFileUrl(normalizedModulePath).href
+
+		const { default: EspeakInitializer } = await import(modulePathAsFileUrl)
 
 		const m = await EspeakInitializer()
 		espeakInstance = await (new m.eSpeakNGWorker())
@@ -592,6 +675,10 @@ async function getEspeakInstance() {
 	return { instance: espeakInstance, module: espeakModule }
 }
 
+export async function ensureEspeakInstanceInitialized(callbacks: OperationCallbacks) {
+	await getEspeakInstance(callbacks)
+}
+
 export type EspeakEventType = 'sentence' | 'word' | 'phoneme' | 'end' | 'mark' | 'play' | 'msg_terminated' | 'list_terminated' | 'samplerate'
 
 export interface EspeakEvent {
@@ -622,8 +709,8 @@ export const defaultEspeakOptions: EspeakOptions = {
 }
 
 export async function testKirshenbaumPhonemization(text: string, language = 'en-us') {
-	const ipaPhonemizedSentence = (await phonemizeSentence(text, language)).flatMap(clause => clause)
-	const kirshenbaumPhonemizedSentence = (await phonemizeSentence(text, language, undefined, false)).flatMap(clause => clause)
+	const ipaPhonemizedSentence = phonemizeSentence(text, language).flatMap(clause => clause)
+	const kirshenbaumPhonemizedSentence = phonemizeSentence(text, language, undefined, false).flatMap(clause => clause)
 
 	const ipaFragments = ipaPhonemizedSentence.map(word => word.join(''))
 

package/src/synthesis/FliteTTS.ts

@@ -1,4 +1,4 @@
-import { SynthesisVoice } from '../api/API.js'
+import { SynthesisCallbacks, SynthesisVoice } from '../api/API.js'
 import { decodeWaveToRawAudio } from '../audio/AudioUtilities.js'
 import { Logger } from '../utilities/Logger.js'
 import { getRandomHexString, logToStderr, resolveModuleMainPath } from '../utilities/Utilities.js'
@@ -11,8 +11,8 @@ export type FliteVoiceName = 'kal' | 'kal16' | 'awb' | 'rms' | 'slt' | string
 
 let fliteModuleObject: WebAssembly.Module
 
-export async function synthesize(text: string, voice: FliteVoiceName, voiceDir: string | undefined, rate: number, pitchMeanHz?: number, pitchStdDev?: number) {
-	const logger = new Logger()
+export async function synthesize(text: string, voice: FliteVoiceName, voiceDir: string | undefined, rate: number, pitchMeanHz: number | undefined, pitchStdDev: number | undefined, callbacks: SynthesisCallbacks) {
+	const logger = new Logger(callbacks.logLevel)
 	logger.start('Get Flite WASI instance')
 
 	const randomId = getRandomHexString(16)

package/src/synthesis/GnuSpeechTTS.ts

@@ -1,9 +1,9 @@
 import { defaultGnuSpeechOptions, GnuSpeechOptions, synthesize as gnuSpeechSynthesize } from '@echogarden/gnuspeech-wasm'
-import { SynthesisVoice } from '../api/Synthesis.js'
+import { SynthesisCallbacks, SynthesisVoice } from '../api/Synthesis.js'
 import { decodeWaveToRawAudio } from '../audio/AudioUtilities.js'
 import { extendDeep } from '../utilities/ObjectUtilities.js'
 
-export async function synthesize(text: string, options: GnuSpeechOptions) {
+export async function synthesize(text: string, options: GnuSpeechOptions, callbacks: SynthesisCallbacks) {
 	const gnuSpeechOptions = extendDeep(defaultGnuSpeechOptions, options)
 
 	const { audioData, params } = await gnuSpeechSynthesize(text, gnuSpeechOptions)

package/src/synthesis/GoogleCloudTTS.ts

@@ -1,8 +1,9 @@
-import { request } from 'gaxios'
 import { Logger } from '../utilities/Logger.js'
 import { logToStderr } from '../utilities/Utilities.js'
 import { decodeBase64 } from '../encodings/Base64.js'
 import * as FFMpegTranscoder from '../codecs/FFMpegTranscoder.js'
+import { requestHttp } from 'easier-http-request'
+import { VoiceListRequestCallbacks, SynthesisCallbacks } from '../api/API.js'
 
 const log = logToStderr
 
@@ -16,9 +17,10 @@ export async function synthesize(
 	volumeGainDecibels = 0.0,
 	ssml = false,
 	audioEncoding: AudioEncoding = 'MP3',
-	sampleRate = 24000) {
+	sampleRate = 24000,
+	callbacks: SynthesisCallbacks) {
 
-	const logger = new Logger()
+	const logger = new Logger(callbacks.logLevel)
 	logger.start('Request synthesis from Google Cloud')
 
 	const requestBody = {
@@ -49,7 +51,7 @@ export async function synthesize(
 		requestBody.input.text = text
 	}
 
-	const response = await request<any>({
+	const response = await requestHttp({
 		method: 'POST',
 
 		url: `https://texttospeech.googleapis.com/v1beta1/text:synthesize`,
@@ -62,18 +64,25 @@ export async function synthesize(
 			'User-Agent': ''
 		},
 
-		data: requestBody,
+		body: requestBody,
 
-		responseType: 'json'
+		abortSignal: callbacks?.abortSignal
 	})
 
+	const responseObject = await response.json()
+
 	logger.start('Parse result')
 
-	const result = parseResponseBody(response.data)
+	const result = parseResponseObject(responseObject)
 
 	logger.start('Decode to raw audio')
 
-	const rawAudio = await FFMpegTranscoder.decodeToChannels(result.audioData)
+	const rawAudio = await FFMpegTranscoder.decodeToChannels(
+		result.audioData,
+		undefined,
+		undefined,
+		{ abortSignal: callbacks.abortSignal, logLevel: 'warning' },
+	)
 	const timepoints = result.timepoints
 
 	logger.end()
@@ -81,7 +90,7 @@ export async function synthesize(
 	return { rawAudio, timepoints }
 }
 
-function parseResponseBody(responseBody: any) {
+function parseResponseObject(responseBody: any) {
 	const audioData = decodeBase64(responseBody.audioContent)
 	const timepoints: timePoint[] = responseBody.timepoints
 
@@ -89,10 +98,10 @@ function parseResponseBody(responseBody: any) {
 }
 
 // Voices with audio samples: https://cloud.google.com/text-to-speech/docs/voices
-export async function getVoiceList(apiKey: string) {
+export async function getVoiceList(apiKey: string, callbacks: VoiceListRequestCallbacks) {
 	const requestURL = `https://texttospeech.googleapis.com/v1beta1/voices`
 
-	const response = await request<any>({
+	const response = await requestHttp({
 		method: 'GET',
 
 		url: requestURL,
@@ -105,12 +114,12 @@ export async function getVoiceList(apiKey: string) {
 			'User-Agent': ''
 		},
 
-		responseType: 'json'
+		abortSignal: callbacks?.abortSignal
 	})
 
-	const responseData = response.data
+	const responseObject = await response.json()
 
-	const voices: GoogleCloudVoice[] = responseData.voices
+	const voices: GoogleCloudVoice[] = responseObject.voices
 
 	return voices
 }

package/src/synthesis/GoogleTranslateTTS.ts

@@ -1,4 +1,3 @@
-import { request } from 'gaxios'
 import { Phrase, splitToFragments } from '../nlp/Segmentation.js'
 import { concatFloat32Arrays, concatUint8Arrays, logToStderr } from '../utilities/Utilities.js'
 import * as FFMpegTranscoder from '../codecs/FFMpegTranscoder.js'
@@ -8,17 +7,18 @@ import { Timeline } from '../utilities/Timeline.js'
 import { getShortLanguageCode } from '../utilities/Locale.js'
 import { getChromeOnWindowsHeaders } from '../utilities/BrowserRequestHeaders.js'
 import { decodeBase64 } from '../encodings/Base64.js'
-
-const log = logToStderr
+import { requestHttp } from 'easier-http-request'
+import { SynthesisCallbacks } from '../api/API.js'
 
 const maxTextLengthPerRequest = 200
 
-export async function synthesizeLongText(text: string, languageCode = 'en', tld = 'us', sentenceEndPause = 0.75, segmentEndPause = 1.0) {
+export async function synthesizeLongText(text: string, languageCode = 'en', tld = 'us', sentenceEndPause = 0.75, segmentEndPause = 1.0, callbacks: SynthesisCallbacks) {
 	if (text.length == 0) {
 		throw new Error('Text is empty')
 	}
 
-	const logger = new Logger()
+	const logger = new Logger(callbacks.logLevel)
+
 	logger.start('Prepare and split text')
 
 	const fragments = await splitToFragments(text, maxTextLengthPerRequest, languageCode)
@@ -32,13 +32,19 @@ export async function synthesizeLongText(text: string, languageCode = 'en', tld
 		const fragment = fragments[i]
 
 		logger.start(`Request synthesis for text fragment ${i + 1}/${fragments.length} from Google Translate`)
-		const fragmentMp3Stream = await synthesizeShortText(fragment.text, languageCode, tld)
+		const fragmentMp3Stream = await synthesizeShortText(fragment.text, languageCode, tld, callbacks)
 
 		if (fragmentMp3Stream.length == 0) {
 			continue
 		}
 
-		const rawAudio = await FFMpegTranscoder.decodeToChannels(fragmentMp3Stream, 24000, 1)
+		const rawAudio = await FFMpegTranscoder.decodeToChannels(
+			fragmentMp3Stream,
+			24000,
+			1,
+			{ abortSignal: callbacks.abortSignal, logLevel: 'warning' }
+		)
+
 		fragmentsSampleRate = rawAudio.sampleRate
 
 		let targetEndingSilenceTime: number
@@ -76,7 +82,7 @@ export async function synthesizeLongText(text: string, languageCode = 'en', tld
 	}
 }
 
-export async function synthesizeShortText(text: string, languageCode = 'en', tld = 'us') {
+export async function synthesizeShortText(text: string, languageCode = 'en', tld = 'us', callbacks: SynthesisCallbacks) {
 	if (text.length > maxTextLengthPerRequest) {
 		throw new Error(`Text is ${text.length} characters, which is longer than the maximum of ${maxTextLengthPerRequest}`)
 	}
@@ -90,7 +96,7 @@ export async function synthesizeShortText(text: string, languageCode = 'en', tld
 	const stringifiedForm = JSON.stringify(requestForm)
 	const requestBody = `f.req=${encodeURIComponent(stringifiedForm)}&`
 
-	const response = await request<string>({
+	const response = await requestHttp({
 		method: 'POST',
 
 		url: `https://translate.google.${tld}/_/TranslateWebserverUi/data/batchexecute`,
@@ -109,14 +115,16 @@ export async function synthesizeShortText(text: string, languageCode = 'en', tld
 
 		body: requestBody,
 
-		responseType: 'text'
+		abortSignal: callbacks?.abortSignal,
 	})
 
+	const responseBody = await response.text()
+
 	//log(response.data)
 
 	const audioChunks: Uint8Array[] = []
 
-	for (const line of response.data.split(/\r?\n/)) {
+	for (const line of responseBody.split(/\r?\n/)) {
 		const match = line.match(/"jQ1olc","\[\\"(.*)\\"]/)
 
 		if (match != null) {