echogarden 2.10.2 → 3.0.1

package/dist/utilities/BinaryUtilities.js

@@ -1,113 +0,0 @@
-/////////////////////////////////////////////////////////////////////////////////////
-// Unsigned operations
-/////////////////////////////////////////////////////////////////////////////////////
-export function readUint8(buffer, offset) {
-    return buffer[offset];
-}
-export function writeUint8(buffer, value, offset) {
-    if (value < 0 || value > 255) {
-        throw new Error(`Value ${value} is outside the range of an 8-bit unsigned integer`);
-    }
-    buffer[offset] = value;
-}
-export function readUint16LE(buffer, offset) {
-    return (buffer[offset]) | (buffer[offset + 1] << 8);
-}
-export function writeUint16LE(buffer, value, offset) {
-    if (value < 0 || value > 65535) {
-        throw new Error(`Value ${value} is outside the range of a 16-bit unsigned integer`);
-    }
-    buffer[offset] = value & 0xff;
-    buffer[offset + 1] = (value >>> 8) & 0xff;
-}
-export function readUint32LE(buffer, offset) {
-    return readInt32LE(buffer, offset) >>> 0;
-}
-export function writeUint32LE(buffer, value, offset) {
-    if (value < 0 || value > 4294967295) {
-        throw new Error(`Value ${value} is outside the range of a 32-bit unsigned integer`);
-    }
-    buffer[offset] = value & 0xff;
-    buffer[offset + 1] = (value >>> 8) & 0xff;
-    buffer[offset + 2] = (value >>> 16) & 0xff;
-    buffer[offset + 3] = (value >>> 24) & 0xff;
-}
-/////////////////////////////////////////////////////////////////////////////////////
-// Signed operations
-/////////////////////////////////////////////////////////////////////////////////////
-export function readInt8(buffer, offset) {
-    const unsignedValue = buffer[offset];
-    if (unsignedValue < 128) {
-        return unsignedValue;
-    }
-    else {
-        return unsignedValue - 256;
-    }
-}
-export function writeInt8(buffer, value, offset) {
-    if (value < -128 || value > 127) {
-        throw new Error(`Value ${value} is outside the range of an 8-bit signed integer`);
-    }
-    let unsignedValue;
-    if (value >= 0) {
-        unsignedValue = value;
-    }
-    else {
-        unsignedValue = value + 256;
-    }
-    buffer[offset] = unsignedValue;
-}
-export function readInt16LE(buffer, offset) {
-    const unsignedValue = readUint16LE(buffer, offset);
-    if (unsignedValue < 32768) {
-        return unsignedValue;
-    }
-    else {
-        return unsignedValue - 65536;
-    }
-}
-export function writeInt16LE(buffer, value, offset) {
-    if (value < -32768 || value > 32767) {
-        throw new Error(`Value ${value} is outside the range of a 16-bit signed integer`);
-    }
-    let unsignedValue;
-    if (value >= 0) {
-        unsignedValue = value;
-    }
-    else {
-        unsignedValue = value + 65536;
-    }
-    writeUint16LE(buffer, unsignedValue, offset);
-}
-export function readInt32LE(buffer, offset) {
-    const value = (buffer[offset]) |
-        (buffer[offset + 1] << 8) |
-        (buffer[offset + 2] << 16) |
-        (buffer[offset + 3] << 24);
-    return value;
-}
-export function writeInt32LE(buffer, value, offset) {
-    if (value < -2147483648 || value > 2147483647) {
-        throw new Error(`Value ${value} is outside the range of a 32-bit signed integer`);
-    }
-    buffer[offset] = value & 0xff;
-    buffer[offset + 1] = (value >> 8) & 0xff;
-    buffer[offset + 2] = (value >> 16) & 0xff;
-    buffer[offset + 3] = (value >> 24) & 0xff;
-}
-/////////////////////////////////////////////////////////////////////////////////////
-// Misc operations
-/////////////////////////////////////////////////////////////////////////////////////
-export function writeAscii(buffer, asciiString, writeStartOffset) {
-    const writeEndOffset = Math.min(writeStartOffset + asciiString.length, buffer.length);
-    let readOffset = 0;
-    let writeOffset = writeStartOffset;
-    while (writeOffset < writeEndOffset) {
-        const charCode = asciiString.charCodeAt(readOffset++);
-        if (charCode >= 128) {
-            throw new Error(`Character '${asciiString[readOffset]}' (code: ${charCode}) can't be encoded as ASCII`);
-        }
-        buffer[writeOffset++] = charCode;
-    }
-}
-//# sourceMappingURL=BinaryUtilities.js.map

package/dist/utilities/BinaryUtilities.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"BinaryUtilities.js","sourceRoot":"","sources":["../../src/utilities/BinaryUtilities.ts"],"names":[],"mappings":"AAAA,qFAAqF;AACrF,sBAAsB;AACtB,qFAAqF;AACrF,MAAM,UAAU,SAAS,CAAC,MAAkB,EAAE,MAAc;IAC3D,OAAO,MAAM,CAAC,MAAM,CAAC,CAAA;AACtB,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC3E,IAAI,KAAK,GAAG,CAAC,IAAI,KAAK,GAAG,GAAG,EAAE,CAAC;QAC9B,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,oDAAoD,CAAC,CAAA;IACpF,CAAC;IAED,MAAM,CAAC,MAAM,CAAC,GAAG,KAAK,CAAA;AACvB,CAAC;AAED,MAAM,UAAU,YAAY,CAAC,MAAkB,EAAE,MAAc;IAC9D,OAAO,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAA;AACpD,CAAC;AAED,MAAM,UAAU,aAAa,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC9E,IAAI,KAAK,GAAG,CAAC,IAAI,KAAK,GAAG,KAAK,EAAE,CAAC;QAChC,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,oDAAoD,CAAC,CAAA;IACpF,CAAC;IAED,MAAM,CAAC,MAAM,CAAC,GAAG,KAAK,GAAG,IAAI,CAAA;IAC7B,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,KAAK,CAAC,CAAC,GAAG,IAAI,CAAA;AAC1C,CAAC;AAED,MAAM,UAAU,YAAY,CAAC,MAAkB,EAAE,MAAc;IAC9D,OAAO,WAAW,CAAC,MAAM,EAAE,MAAM,CAAC,KAAK,CAAC,CAAA;AACzC,CAAC;AAED,MAAM,UAAU,aAAa,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC9E,IAAI,KAAK,GAAG,CAAC,IAAI,KAAK,GAAG,UAAU,EAAE,CAAC;QACrC,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,oDAAoD,CAAC,CAAA;IACpF,CAAC;IAED,MAAM,CAAC,MAAM,CAAC,GAAG,KAAK,GAAG,IAAI,CAAA;IAC7B,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,KAAK,CAAC,CAAC,GAAG,IAAI,CAAA;IACzC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,KAAK,EAAE,CAAC,GAAG,IAAI,CAAA;IAC1C,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,KAAK,EAAE,CAAC,GAAG,IAAI,CAAA;AAC3C,CAAC;AAED,qFAAqF;AACrF,oBAAoB;AACpB,qFAAqF;AACrF,MAAM,UAAU,QAAQ,CAAC,MAAkB,EAAE,MAAc;IAC1D,MAAM,aAAa,GAAG,MAAM,CAAC,MAAM,CAAC,CAAA;IAEpC,IAAI,aAAa,GAAG,GAAG,EAAE,CAAC;QACzB,OAAO,aAAa,CAAA;IACrB,CAAC;SAAM,CAAC;QACP,OAAO,aAAa,GAAG,GAAG,CAAA;IAC3B,CAAC;AACF,CAAC;AAED,MAAM,UAAU,SAAS,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC1E,IAAI,KAAK,GAAG,CAAC,GAAG,IAAI,KAAK,GAAG,GAAG,EAAE,CAAC;QACjC,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,kDAAkD,CAAC,CAAA;IAClF,CAAC;IAED,IAAI,aAAqB,CAAA;IAEzB,IAAI,KAAK,IAAI,CAAC,EAAE,CAAC;QAChB,aAAa,GAAG,KAAK,CAAA;IACtB,CAAC;SAAM,CAAC;QACP,aAAa,GAAG,KAAK,GAAG,GAAG,CAAA;IAC5B,CAAC;IAED,MAAM,CAAC,MAAM,CAAC,GAAG,aAAa,CAAA;AAC/B,CAAC;AAED,MAAM,UAAU,WAAW,CAAC,MAAkB,EAAE,MAAc;IAC7D,MAAM,aAAa,GAAG,YAAY,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;IAElD,IAAI,aAAa,GAAG,KAAK,EAAE,CAAC;QAC3B,OAAO,aAAa,CAAA;IACrB,CAAC;SAAM,CAAC;QACP,OAAO,aAAa,GAAG,KAAK,CAAA;IAC7B,CAAC;AACF,CAAC;AAED,MAAM,UAAU,YAAY,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC7E,IAAI,KAAK,GAAG,CAAC,KAAK,IAAI,KAAK,GAAG,KAAK,EAAE,CAAC;QACrC,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,kDAAkD,CAAC,CAAA;IAClF,CAAC;IAED,IAAI,aAAqB,CAAA;IAEzB,IAAI,KAAK,IAAI,CAAC,EAAE,CAAC;QAChB,aAAa,GAAG,KAAK,CAAA;IACtB,CAAC;SAAM,CAAC;QACP,aAAa,GAAG,KAAK,GAAG,KAAK,CAAA;IAC9B,CAAC;IAED,aAAa,CAAC,MAAM,EAAE,aAAa,EAAE,MAAM,CAAC,CAAA;AAC7C,CAAC;AAED,MAAM,UAAU,WAAW,CAAC,MAAkB,EAAE,MAAc;IAC7D,MAAM,KAAK,GACV,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;QAChB,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC;QACzB,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;QAC1B,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,CAAA;IAE3B,OAAO,KAAK,CAAA;AACb,CAAC;AAED,MAAM,UAAU,YAAY,CAAC,MAAkB,EAAE,KAAa,EAAE,MAAc;IAC7E,IAAI,KAAK,GAAG,CAAC,UAAU,IAAI,KAAK,GAAG,UAAU,EAAE,CAAC;QAC/C,MAAM,IAAI,KAAK,CAAC,SAAS,KAAK,kDAAkD,CAAC,CAAA;IAClF,CAAC;IAED,MAAM,CAAC,MAAM,CAAC,GAAG,KAAK,GAAG,IAAI,CAAA;IAC7B,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,IAAI,CAAC,CAAC,GAAG,IAAI,CAAA;IACxC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,GAAG,IAAI,CAAA;IACzC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC,GAAG,IAAI,CAAA;AAC1C,CAAC;AAED,qFAAqF;AACrF,kBAAkB;AAClB,qFAAqF;AACrF,MAAM,UAAU,UAAU,CAAC,MAAkB,EAAE,WAAmB,EAAE,gBAAwB;IAC3F,MAAM,cAAc,GAAG,IAAI,CAAC,GAAG,CAAC,gBAAgB,GAAG,WAAW,CAAC,MAAM,EAAE,MAAM,CAAC,MAAM,CAAC,CAAA;IAErF,IAAI,UAAU,GAAG,CAAC,CAAA;IAClB,IAAI,WAAW,GAAG,gBAAgB,CAAA;IAElC,OAAO,WAAW,GAAG,cAAc,EAAE,CAAC;QACrC,MAAM,QAAQ,GAAG,WAAW,CAAC,UAAU,CAAC,UAAU,EAAE,CAAC,CAAA;QAErD,IAAI,QAAQ,IAAI,GAAG,EAAE,CAAC;YACrB,MAAM,IAAI,KAAK,CAAC,cAAc,WAAW,CAAC,UAAU,CAAC,YAAY,QAAQ,6BAA6B,CAAC,CAAA;QACxG,CAAC;QAED,MAAM,CAAC,WAAW,EAAE,CAAC,GAAG,QAAQ,CAAA;IACjC,CAAC;AACF,CAAC"}

package/src/audio/AudioRecorder.ts

@@ -1,91 +0,0 @@
-import { spawn } from 'child_process'
-
-import { RawAudio, concatAudioSegments } from './AudioUtilities.js'
-import * as AudioBufferConversion from './AudioBufferConversion.js'
-
-import { Timer } from '../utilities/Timer.js'
-import { logToStderr } from '../utilities/Utilities.js'
-import { OpenPromise } from '../utilities/OpenPromise.js'
-import { tryResolvingSoxPath } from './SoxPath.js'
-import { SampleFormat } from './AudioBufferConversion.js'
-
-const log = logToStderr
-
-export async function recordAudioInput(channelCount = 1, sampleRate = 48000, maxTime = 10) {
-	const audioSegments: Float32Array[][] = []
-
-	await captureAudioInput(channelCount, sampleRate, maxTime, (rawAudio) => {
-		audioSegments.push(rawAudio)
-	})
-
-	const mergedAudio = concatAudioSegments(audioSegments)
-
-	const rawAudio: RawAudio = {
-		audioChannels: mergedAudio,
-		sampleRate
-	}
-
-	return rawAudio
-}
-
-export function captureAudioInput(channelCount = 1, sampleRate = 48000, maxTime = -1, onAudioSamples?: (rawAudio: Float32Array[]) => void) {
-	return new Promise<void>(async (resolve, reject) => {
-		const timer = new Timer()
-
-		const recorderSpawnedOpenPromise = new OpenPromise<null>()
-
-		const soxPath = await tryResolvingSoxPath()
-
-		if (!soxPath) {
-			throw new Error('Could not resolve a SoX executable')
-		}
-
-		let args: string[]
-
-		if (process.platform == 'win32') {
-			args = ['-b', `${16}`, '--endian', 'little',
-				'-c', `${channelCount}`, '-r', `${sampleRate}`, '-e', 'signed',
-				'-t', 'waveaudio', 'default', '-t', 'raw', '-']
-		} else if (process.platform == 'darwin') {
-			args = ['-b', `${16}`, '--endian', 'little',
-				'-c', `${channelCount}`, '-r', `${sampleRate}`, '-e', 'signed',
-				'-t', 'raw', '-']
-		} else {
-			throw new Error('')
-		}
-
-		const recorder = spawn(
-			soxPath,
-			args,
-			{}
-		)
-
-		recorder.once('spawn', () => {
-			recorderSpawnedOpenPromise.resolve(null)
-			timer.restart()
-		})
-
-		recorder.once('close', () => {
-			resolve()
-		})
-
-		recorder.stderr.on('data', (chunk) => {
-			//log(chunk.toString('utf8'))
-		})
-
-		recorder.stdout.on('data', (chunk) => {
-			//log(chunk)
-
-			const audioChannels = AudioBufferConversion.decodeToChannels(chunk, channelCount, 16, SampleFormat.PCM)
-
-			if (onAudioSamples) {
-				onAudioSamples(audioChannels)
-			}
-
-			if (maxTime >= 0 && timer.elapsedTimeSeconds >= maxTime) {
-				recorder.kill()
-				resolve()
-			}
-		})
-	})
-}

package/src/audio/SoxPath.ts

@@ -1,31 +0,0 @@
-import { loadPackage } from '../utilities/PackageManager.js'
-import { commandExists } from '../utilities/Utilities.js'
-import { getGlobalOption } from '../api/GlobalOptions.js'
-import { joinPath } from '../utilities/PathUtilities.js'
-
-export async function tryResolvingSoxPath() {
-	if (getGlobalOption('soxPath')) {
-		return getGlobalOption('soxPath')
-	}
-
-	const platform = process.platform
-	const arch = process.arch
-
-	if (platform === 'win32') {
-		const soxPackagePath = await loadPackage('sox-14.4.1a-win32')
-
-		return joinPath(soxPackagePath, 'sox.exe')
-	}
-
-	if (await commandExists('sox')) {
-		return 'sox'
-	}
-
-	if (platform === 'linux' && arch === 'x64') {
-		const soxPackagePath = await loadPackage('sox-14.4.2-linux-minimal')
-
-		return joinPath(soxPackagePath, 'sox')
-	}
-
-	return undefined
-}

package/src/recognition/SileroSTT.ts

@@ -1,222 +0,0 @@
-import { indexOfMax } from '../math/VectorMath.js'
-import { wordCharacterRegExp } from '../nlp/Segmentation.js'
-import { Logger } from '../utilities/Logger.js'
-import { logToStderr } from '../utilities/Utilities.js'
-import { Timeline } from '../utilities/Timeline.js'
-import { RawAudio, getRawAudioDuration } from '../audio/AudioUtilities.js'
-import { readAndParseJsonFile } from '../utilities/FileSystem.js'
-
-import type * as Onnx from 'onnxruntime-node'
-import { OnnxExecutionProvider, getOnnxSessionOptions } from '../utilities/OnnxUtilities.js'
-import { joinPath } from '../utilities/PathUtilities.js'
-
-const log = logToStderr
-
-export async function recognize(
-	rawAudio: RawAudio,
-	modelDirectoryPath: string,
-	executionProviders: OnnxExecutionProvider[]) {
-
-	const silero = new SileroSTT(modelDirectoryPath, executionProviders)
-
-	const result = await silero.recognize(rawAudio)
-
-	return result
-}
-
-export class SileroSTT {
-	session?: Onnx.InferenceSession
-	labels?: string[]
-
-	constructor(
-		public readonly modelDirectoryPath: string,
-		public readonly executionProviders: OnnxExecutionProvider[]
-	) {
-	}
-
-	async recognize(rawAudio: RawAudio) {
-		const logger = new Logger()
-
-		await this.initializeIfNeeded()
-
-		logger.start('Recognize with silero model')
-
-		const audioSamples = rawAudio.audioChannels[0]
-
-		const Onnx = await import('onnxruntime-node')
-
-		const inputTensor = new Onnx.Tensor('float32', audioSamples, [1, audioSamples.length])
-		const inputs = { input: inputTensor }
-
-		const results = await this.session!.run(inputs)
-
-		const rawResultValues = results['output'].data as Float32Array
-
-		const labels = this.labels!
-
-		const tokenResults: Float32Array[] = []
-
-		for (let i = 0; i < rawResultValues.length; i += labels.length) {
-			tokenResults.push(rawResultValues.subarray(i, i + labels.length))
-		}
-
-		const tokens: string[] = []
-
-		for (const tokenResult of tokenResults) {
-			const bestCandidateIndex = indexOfMax(new Array(...tokenResult))
-			tokens.push(labels[bestCandidateIndex])
-		}
-
-		//log(tokens.join('|'))
-
-		const result = this.tokensToTimeline(tokens, getRawAudioDuration(rawAudio))
-
-		logger.end()
-
-		return result
-	}
-
-	private async initializeIfNeeded() {
-		if (this.session) {
-			return
-		}
-
-		const logger = new Logger()
-
-		logger.start('Create ONNX inference session')
-
-		const Onnx = await import('onnxruntime-node')
-
-		const modelPath = joinPath(this.modelDirectoryPath, 'model.onnx')
-		const labelsPath = joinPath(this.modelDirectoryPath, 'labels.json')
-
-		this.labels = await readAndParseJsonFile(labelsPath)
-
-		const onnxSessionOptions = getOnnxSessionOptions({ executionProviders: this.executionProviders })
-
-		this.session = await Onnx.InferenceSession.create(modelPath, onnxSessionOptions)
-
-		logger.end()
-	}
-
-	private tokensToTimeline(tokens: string[], totalDuration: number) {
-		const tokenCount = tokens.length
-
-		const decodedTokens: string[] = []
-		let tokenGroupIndexes: number[][] = [[]]
-
-		for (let i = 0; i < tokenCount; i++) {
-			const token = tokens[i]
-
-			if (token == '2') {
-				if (decodedTokens.length > 0) {
-					const previousDecodedToken = decodedTokens[decodedTokens.length - 1]
-					decodedTokens.push('$')
-					decodedTokens.push(previousDecodedToken)
-
-					tokenGroupIndexes[tokenGroupIndexes.length - 1].push(i)
-				} else {
-					decodedTokens.push(' ')
-					tokenGroupIndexes.push([])
-				}
-
-				continue
-			}
-
-			if (token == '_') {
-				continue
-			}
-
-			decodedTokens.push(token)
-
-			if (token == ' ') {
-				tokenGroupIndexes.push([])
-			} else {
-				tokenGroupIndexes[tokenGroupIndexes.length - 1].push(i)
-			}
-		}
-
-		let decodedString = ''
-
-		for (let i = 0; i < decodedTokens.length; i++) {
-			const currentToken = decodedTokens[i]
-			const previousToken = decodedTokens[i - 1]
-
-			if (currentToken != '$' && (previousToken != currentToken || previousToken == undefined)) {
-				decodedString += currentToken
-			}
-		}
-
-		decodedString = decodedString.trim()
-
-		tokenGroupIndexes = tokenGroupIndexes.filter(group => group.length > 0)
-
-		if (tokenGroupIndexes.length > 0) {
-			let currentCorrection = Math.min(tokenGroupIndexes[0][0], 1.5)
-
-			for (let i = 0; i < tokenGroupIndexes.length; i++) {
-				const group = tokenGroupIndexes[i]
-
-				if (group.length == 1) {
-					group.push(group[0])
-				}
-
-				group[0] -= currentCorrection
-
-				if (i == tokenGroupIndexes.length - 1) {
-					currentCorrection = Math.min(tokenCount - i, 1.5)
-				} else {
-					currentCorrection = Math.min((tokenGroupIndexes[i + 1][0] - group[group.length - 1]) / 2, 1.5)
-				}
-
-				group[group.length - 1] += currentCorrection
-			}
-		}
-
-		const words = decodedString.split(' ')
-
-		const timeMultiplier = totalDuration / tokenCount
-
-		const timeline: Timeline = []
-
-		for (let i = 0; i < words.length; i++) {
-			const text = words[i]
-
-			if (!wordCharacterRegExp.test(text)) {
-				continue
-			}
-
-			const group = tokenGroupIndexes[i]
-			const startTime = group[0] * timeMultiplier
-			const endTime = group[group.length - 1] * timeMultiplier
-
-			timeline.push({
-				type: 'word',
-				text: text,
-				startTime,
-				endTime,
-			})
-		}
-
-		timeline[timeline.length - 1].endTime = totalDuration
-
-		return { transcript: decodedString, timeline }
-	}
-}
-
-export const languageCodeToPackageName: { [languageCode: string]: string } = {
-	'en': 'silero-en-v5',
-	'es': 'silero-es-v1',
-	'de': 'silero-de-v1',
-	'uk': 'silero-ua-v3',
-}
-
-export interface SileroRecognitionOptions {
-	modelPath?: string
-	provider?: OnnxExecutionProvider
-}
-
-export const defaultSileroRecognitionOptions: SileroRecognitionOptions = {
-	modelPath: undefined,
-	provider: undefined,
-}

package/src/recognition/VoskSTT.ts

@@ -1,112 +0,0 @@
-import * as FFMpegTranscoder from '../codecs/FFMpegTranscoder.js'
-import * as AudioBufferConversion from '../audio/AudioBufferConversion.js'
-import { Logger } from '../utilities/Logger.js'
-import { concatUint8Arrays, logToStderr } from '../utilities/Utilities.js'
-import { Timeline } from '../utilities/Timeline.js'
-import { RawAudio } from '../audio/AudioUtilities.js'
-
-const log = logToStderr
-
-export async function recognizeFile(filename: string, modelPath: string, verbose = true) {
-	const rawAudio = await FFMpegTranscoder.decodeToChannels(filename, 16000, 1)
-	return recognize(rawAudio, modelPath, verbose)
-}
-
-export async function recognize(rawAudio: RawAudio, modelPath: string, verbose = true) {
-	const logger = new Logger()
-	logger.start('Initialize vosk recognizer')
-
-	const audioChannels = rawAudio.audioChannels
-	const sampleRate = rawAudio.sampleRate
-
-	let Vosk = await import('@echogarden/vosk')
-
-	Vosk.setLogLevel(-1)
-
-	const model = await new Vosk.Model(modelPath)
-
-	const recognizer = new Vosk.Recognizer({ model, sampleRate })
-
-	recognizer.setMaxAlternatives(0)
-	recognizer.setWords(true)
-	recognizer.setPartialWords(true)
-
-	logger.start('Recognize with vosk')
-
-	const recognitionStartTimestamp = logger.getTimestamp()
-
-	const pcmAudio = AudioBufferConversion.encodeToAudioBuffer(audioChannels, 16)
-	const trailingSilence = new Uint8Array(sampleRate * 4)
-	const pcmAudioWithTrailingSilence = concatUint8Arrays([pcmAudio, trailingSilence])
-	const pcmAudioByteCount = pcmAudioWithTrailingSilence.length
-
-	const maxChunkSize = sampleRate * 2.0
-
-	let previousResultText = ''
-
-	for (let readOffset = 0; readOffset < pcmAudioByteCount; readOffset += maxChunkSize) {
-		const chunkSize = Math.min(maxChunkSize, pcmAudioByteCount - readOffset)
-
-		const chunk = pcmAudioWithTrailingSilence.subarray(readOffset, readOffset + chunkSize)
-
-		const speechEnded = await recognizer.acceptWaveformAsync(chunk)
-
-		if (verbose) {
-			const partialResultText = recognizer.partialResult().partial
-
-			if (partialResultText != previousResultText) {
-				//logger.log(partialResultText)
-				//logger.log('')
-
-				previousResultText = partialResultText
-			}
-		}
-	}
-
-	//const speechEnded = await recognizer.acceptWaveformAsync(pcmAudioWithTrailingSilence)
-
-	const result = recognizer.finalResult()
-
-	recognizer.reset()
-	recognizer.free()
-	model.free()
-
-	const transcript: string = result.text
-	const events: VoskRecognitionEvent[] = result.result
-
-	if (events.length == 0) {
-		return { transcript, timeline: [] }
-	}
-
-	const timeline: Timeline = []
-
-	for (let i = 0; i < events.length; i++) {
-		const event = events[i]
-
-		const eventText = event.word
-		const eventStart = event.start
-		const eventEnd = event.end
-		const eventConfidence = event.conf
-
-		timeline.push({
-			type: 'word',
-			text: eventText,
-			startTime: eventStart,
-			endTime: eventEnd,
-			confidence: eventConfidence
-		})
-	}
-
-	//logger.logDuration(`Recognition with vosk`, recognitionStartTimestamp)
-
-	logger.end()
-
-	return { transcript, timeline }
-}
-
-type VoskRecognitionEvent = {
-	word: string
-	start: number
-	end: number
-	conf: number
-}

package/src/synthesis/StreamlabsPollyTTS.ts

@@ -1,114 +0,0 @@
-import { request } from 'gaxios'
-import { SynthesisVoice } from '../api/API.js'
-import { trimAudioEnd } from '../audio/AudioUtilities.js'
-import * as FFMpegTranscoder from '../codecs/FFMpegTranscoder.js'
-import { Phrase, splitToFragments } from '../nlp/Segmentation.js'
-import { Logger } from '../utilities/Logger.js'
-import { concatFloat32Arrays, logToStderr } from '../utilities/Utilities.js'
-import { Timeline } from '../utilities/Timeline.js'
-
-const log = logToStderr
-
-const maxTextLengthPerRequest = 200
-
-export async function synthesizeLongText(text: string, voice: string, languageCode: string, sentenceEndPause = 0.75, segmentEndPause = 1.0) {
-	if (text.length == 0) {
-		throw new Error('Text is empty')
-	}
-
-	const logger = new Logger()
-	logger.start('Prepare and split text')
-
-	const fragments = await splitToFragments(text, maxTextLengthPerRequest, languageCode)
-
-	const audioFragments: Float32Array[] = []
-	let fragmentsSampleRate = 0
-
-	const timeline: Timeline = []
-
-	for (let i = 0; i < fragments.length; i++) {
-		const fragment = fragments[i]
-
-		logger.start(`Request synthesis for text fragment ${i + 1}/${fragments.length} from Streamslabs Polly`)
-		const fragmentMp3Stream = await synthesizeFragment(fragment.text, voice)
-
-		if (fragmentMp3Stream.length == 0) {
-			continue
-		}
-
-		const rawAudio = await FFMpegTranscoder.decodeToChannels(fragmentMp3Stream, 24000, 1)
-		fragmentsSampleRate = rawAudio.sampleRate
-
-		let targetEndingSilenceTime: number
-
-		if (fragment.lastSegment?.isSentenceFinalizer) {
-			targetEndingSilenceTime = sentenceEndPause
-		} else if (fragment.lastSegment instanceof Phrase) {
-			targetEndingSilenceTime = segmentEndPause
-		} else {
-			targetEndingSilenceTime = 0
-		}
-
-		const trimmedAudio = trimAudioEnd(rawAudio.audioChannels[0], targetEndingSilenceTime * rawAudio.sampleRate)
-
-		audioFragments.push(trimmedAudio)
-
-		const startTime = timeline.length == 0 ? 0 : timeline[timeline.length - 1].endTime
-		const endTime = startTime + (trimmedAudio.length / fragmentsSampleRate)
-
-		timeline.push({
-			type: 'segment',
-			text: fragment.text,
-			startTime,
-			endTime
-		})
-	}
-
-	const rawAudio = { audioChannels: [concatFloat32Arrays(audioFragments)], sampleRate: fragmentsSampleRate }
-
-	logger.end()
-
-	return {
-		rawAudio,
-		timeline
-	}
-}
-
-export async function synthesizeFragment(text: string, voice: string) {
-	const response = await request<any>({
-		url: `https://streamlabs.com/polly/speak`,
-
-		method: 'POST',
-
-		data: {
-			voice,
-			text,
-		},
-
-		responseType: 'json'
-	})
-
-	const responseObject = response.data
-	const audioUrl = responseObject.speak_url
-	const audioUrlResponse = await request<ArrayBuffer>({ url: audioUrl, responseType: 'arraybuffer' })
-
-	return new Uint8Array(audioUrlResponse.data)
-}
-
-export const voiceList: SynthesisVoice[] = [
-	{ name: 'Brian', languages: ['en-GB', 'en'], gender: 'male' },
-	{ name: 'Emma', languages: ['en-GB', 'en'], gender: 'female' },
-	{ name: 'Russell', languages: ['en-AU', 'en'], gender: 'male' },
-	{ name: 'Joey', languages: ['en-US', 'en'], gender: 'male' },
-	{ name: 'Matthew', languages: ['en-US', 'en'], gender: 'male' },
-	{ name: 'Joanna', languages: ['en-US', 'en'], gender: 'female' },
-	{ name: 'Kimberly', languages: ['en-US', 'en'], gender: 'female' },
-	{ name: 'Amy', languages: ['en-GB', 'en'], gender: 'female' },
-	{ name: 'Geraint', languages: ['en-GB-WLS', 'en-GB', 'en'], gender: 'male' },
-	{ name: 'Nicole', languages: ['en-AU', 'en'], gender: 'female' },
-	{ name: 'Justin', languages: ['en-US', 'en'], gender: 'male' },
-	{ name: 'Ivy', languages: ['en-US', 'en'], gender: 'female' },
-	{ name: 'Kendra', languages: ['en-US', 'en'], gender: 'female' },
-	{ name: 'Salli', languages: ['en-US', 'en'], gender: 'female' },
-	{ name: 'Raveena', languages: ['en-IN', 'en'], gender: 'female' },
-]