@omnimedia/omnitool 1.1.0-5 → 1.1.0-51

package/s/tools/speech-recognition/common/model.ts

@@ -1,26 +0,0 @@
-import {pub} from "@e280/stz"
-
-import {Loader} from "../../common/loader.js"
-import {DecoderSource} from "../../../driver/fns/schematic.js"
-import {SpeechRecognizerModels, Word, WordGroup} from "../whisper/parts/types.js"
-
-export abstract class SpeechRecognizer extends Loader {
-	multilingual = true
-
-	static speechRecognizerEvents = {
-		onTranscriptionChunk: pub<Word[]>(),
-		onTranscribeProgress: pub<[number]>()
-	}
-
-	abstract transcribe(input: DecoderSource): Promise<WordGroup>
-
-	setMultilingual(value: boolean) {
-		this.multilingual = value
-	}
-
-	detectLanguage?(input: Blob | AudioBuffer): Promise<string>
-
-	setModel(value: SpeechRecognizerModels) {
-		this.model = value
-	}
-}

package/s/tools/speech-recognition/whisper/fns/host.ts

@@ -1,25 +0,0 @@
-
-import {Comrade} from "@e280/comrade"
-import {ProgressItem} from "../parts/types.js"
-import {SpeechRecognizerHostEvents, WhisperSchematic} from "./schematic.js"
-
-export const setupWhisperHost = (events: SpeechRecognizerHostEvents) => (
-	Comrade.host<WhisperSchematic>(_shell => ({
-		async updateModelLoadProgress(item) {
-			events.onModelLoadProgress.pub(item)
-		},
-		async deliverTranscriptionChunk(chunk) {
-			events.onTranscriptionChunk.pub({
-				text: chunk.text,
-				timestamp: chunk.timestamp
-			})
-		},
-		async updateTps(value) {
-			events.onTpsUpdate.pub(value)
-		},
-		async updateTranscribeProgress(value) {
-			events.onTranscribeProgress(value)
-		}
-	}))
-)
-

package/s/tools/speech-recognition/whisper/fns/schematic.ts

@@ -1,23 +0,0 @@
-import {Pub} from "@e280/stz"
-import {AsSchematic} from "@e280/comrade"
-
-import {LoaderEvents} from "../../../common/loader.js"
-import {ProgressItem, TranscriptionChunk, TranscriptionMessage, TranscriptionResult, Word} from "../parts/types.js"
-
-export type WhisperSchematic = AsSchematic<{
-	work: {
-		transcribe(input: TranscriptionMessage): Promise<TranscriptionResult | null>
-	},
-
-	host: {
-		updateModelLoadProgress(item: ProgressItem): Promise<void>
-		deliverTranscriptionChunk(chunk: TranscriptionChunk): Promise<void>
-		updateTps(value: number): Promise<void>
-		updateTranscribeProgress(value: number): Promise<void>
-	}
-}>
-
-export interface SpeechRecognizerHostEvents extends LoaderEvents {
-	onTranscriptionChunk: Pub<Word[]>
-	onTranscribeProgress: Pub<[number]>
-}

package/s/tools/speech-recognition/whisper/fns/work.ts

@@ -1,91 +0,0 @@
-import {Comrade} from "@e280/comrade"
-//@ts-ignore
-import {pipeline, WhisperTextStreamer} from "https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.7.0/dist/transformers.min.js"
-
-import {WhisperSchematic} from "./schematic.js"
-import {TranscriptionChunk} from "../parts/types.js"
-import {PipelineFactory} from "../../../common/transformer-pipeline.js"
-
-// TODO suspicious globals, probably bad
-const pipeline = new PipelineFactory("automatic-speech-recognition")
-let transcriber: any
-
-export const setupWhisperWork = Comrade.work<WhisperSchematic>(shell => ({
-	async transcribe({audio, model, language, duration}) {
-		const isDistil = model.startsWith("distil-whisper/")
-
-		if(!pipeline.model || pipeline.model !== model) {
-			pipeline.instance?.dispose()?.()
-			pipeline.instance = null
-			transcriber = await pipeline.createInstance(
-				model,
-				(data) => {
-					if(data.progress)
-						shell.host.updateModelLoadProgress({
-							id: data.file,
-							progress: data.progress
-						})
-				}
-			)
-		}
-
-		const timePrecision =
-			transcriber.processor.feature_extractor.config.chunk_length /
-			transcriber.model.config.max_source_positions
-
-		const chunkLength = isDistil ? 20 : 30
-		const strideLength = isDistil ? 3 : 5
-
-		let chunkCount = 0
-		let startTime: number | null = null
-		let tokenCount = 0
-		let tps = 0
-
-		const chunkDuration = chunkLength - strideLength
-
-		const estimateProgress = () => {
-			const audioProgressSeconds = chunkCount * chunkDuration
-			return Math.min(audioProgressSeconds / duration, 1)
-		}
-
-		const streamer = new WhisperTextStreamer(transcriber.tokenizer, {
-			time_precision: timePrecision,
-			token_callback_function: () => {
-				startTime ??= performance.now()
-				if (++tokenCount > 1) {
-					tps = (tokenCount / (performance.now() - startTime)) * 1000
-					shell.host.updateTps(tps)
-				}
-			},
-			callback_function: (textChunk: any) => {
-				shell.host.deliverTranscriptionChunk(textChunk)
-			},
-			on_finalize: () => {
-				startTime = null
-				tokenCount = 0
-				chunkCount++
-				const progress = estimateProgress()
-				shell.host.updateTranscribeProgress(progress)
-			},
-		})
-
-		const output = await transcriber(audio, {
-			top_k: 0,
-			do_sample: false,
-			chunk_length_s: chunkLength,
-			stride_length_s: strideLength,
-			language,
-			task: "transcribe",
-			return_timestamps: "word", // if using "word" the on_chunk_start & end is not called thus we cant retrieve timestamps, only after whole thing finishes
-			force_full_sequences: false,
-			streamer,
-		})
-
-		if (!output) return null
-
-		return {
-			tps,
-			...output,
-		}
-	}
-}))

package/s/tools/speech-recognition/whisper/parts/types.ts

@@ -1,38 +0,0 @@
-export interface ProgressItem {
-	id: string
-	progress: number
-}
-
-export type Word = {
-	text: string
-	timestamp: [start: number, end: number]
-}
-
-export type WordGroup = Word[]
-export type Transcript = WordGroup[]
-
-export interface TranscriptionChunk {
-	text: string
-	offset: number
-	timestamp: [number, number]
-	finalised: boolean
-}
-
-export interface TranscriptionMessage {
-	audio: Float32Array
-	model: string
-	subtask: string | null
-	language: string | null
-	duration: number
-}
-
-export interface TranscriptionResult {
-	text: string
-	chunks: TranscriptionChunk[]
-	tps: number
-}
-
-export type ProgressCallback = (data: any) => void
-
-export type SpeechRecognizerModels = "onnx-community/whisper-tiny_timestamped"
-export type SpeechRecognizerSubtasks = "transcribe"

package/s/tools/speech-recognition/whisper/parts/worker.bundle.ts

@@ -1,7 +0,0 @@
-import {Comrade} from "@e280/comrade"
-
-import {setupWhisperWork} from "../fns/work.js"
-import {WhisperSchematic} from "../fns/schematic.js"
-
-await Comrade.worker<WhisperSchematic>(setupWhisperWork)
-

package/s/tools/speech-recognition/whisper/tool.ts

@@ -1,70 +0,0 @@
-import {Comrade, Thread} from "@e280/comrade"
-
-import {WordGroup} from "./parts/types.js"
-import {context} from "../../../context.js"
-import {setupWhisperHost} from "./fns/host.js"
-import {SpeechRecognizer} from "../common/model.js"
-import {WhisperSchematic} from "./fns/schematic.js"
-
-export class Whisper extends SpeechRecognizer {
-	constructor(public thread: Thread<WhisperSchematic>) {
-		super('whisper', "onnx-community/whisper-tiny_timestamped")
-	}
-
-	static async setup() {
-		const thread = await Comrade.thread<WhisperSchematic>({
-			label: "OmnitoolDriver",
-			workerUrl: new URL("/tools/speech-recognition/whisper/parts/worker.bundle.min.js", import.meta.url),
-			setupHost: setupWhisperHost({
-				...this.loaderEvents,
-				...this.speechRecognizerEvents
-			})
-		})
-		return new this(thread)
-	}
-
-	async init() {
-		// there should be called loading of the model in worker instead when transcribe is called ..
-	}
-
-	async #transcribe(source: Blob, options?: {multilingual?: boolean, language?: string}) {
-		const arrayBuffer = await source.arrayBuffer()
-		const audioCTX = new AudioContext({sampleRate: 16000})
-		const audioData = await audioCTX.decodeAudioData(arrayBuffer)
-		let audio
-		if (audioData.numberOfChannels === 2) {
-			const SCALING_FACTOR = Math.sqrt(2)
-			const left = audioData.getChannelData(0)
-			const right = audioData.getChannelData(1)
-			audio = new Float32Array(left.length)
-			for (let i = 0; i < audioData.length; ++i) {
-				audio[i] = (SCALING_FACTOR * (left[i] + right[i])) / 2
-			}
-		} else {
-			audio = audioData.getChannelData(0)
-		}
-		const driver = await context.driver
-		const duration = await driver.getAudioDuration(source)
-		return await this.thread.work.transcribe({
-			audio,
-			duration,
-			model: this.model,
-			subtask: this.multilingual ? "transcribe" : null,
-			language:
-				this.multilingual && options?.language !== "auto"
-					? options?.language ?? "english"
-					: null
-		})
-	}
-
-	async transcribe(source: Blob): Promise<WordGroup> {
-		const result = await this.#transcribe(source)
-
-		const words = result?.chunks.map((chunk: any) => ({
-			text: chunk.text.trim(),
-			timestamp: chunk.timestamp,
-		})) as WordGroup
-
-		return words
-	}
-}

package/x/tests.test.js

@@ -1,6 +0,0 @@
-import { Science } from "@e280/science";
-import driver from "./driver/driver.test.js";
-await Science.run({
-    driver,
-});
-//# sourceMappingURL=tests.test.js.map

package/x/tests.test.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"tests.test.js","sourceRoot":"","sources":["../s/tests.test.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AACrC,OAAO,MAAM,MAAM,yBAAyB,CAAA;AAE5C,MAAM,OAAO,CAAC,GAAG,CAAC;IACjB,MAAM;CACN,CAAC,CAAA"}

package/x/timeline/sugar/omni-test.d.ts

@@ -1 +0,0 @@
-export {};

package/x/timeline/sugar/omni-test.js

@@ -1,22 +0,0 @@
-import { Omni } from "./omni.js";
-import { dummyData } from "../utils/dummy-data.js";
-//
-// create an omni context
-//
-const omni = new Omni();
-//
-// load in some media resources
-//
-const { mediaA, mediaB } = await omni.load({
-    mediaA: dummyData(),
-    mediaB: dummyData(),
-});
-//
-// create a timeline
-//
-const timeline = omni.timeline(o => o.sequence(o.clip(mediaA), o.transition.crossfade(600), o.stack(o.clip(mediaB), o.text("hello world"))));
-//
-// log the timeline
-//
-console.log(JSON.stringify(timeline, undefined, "  "));
-//# sourceMappingURL=omni-test.js.map

package/x/timeline/sugar/omni-test.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"omni-test.js","sourceRoot":"","sources":["../../../s/timeline/sugar/omni-test.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,IAAI,EAAC,MAAM,WAAW,CAAA;AAC9B,OAAO,EAAC,SAAS,EAAC,MAAM,wBAAwB,CAAA;AAEhD,EAAE;AACF,yBAAyB;AACzB,EAAE;AAEF,MAAM,IAAI,GAAG,IAAI,IAAI,EAAE,CAAA;AAEvB,EAAE;AACF,+BAA+B;AAC/B,EAAE;AAEF,MAAM,EAAC,MAAM,EAAE,MAAM,EAAC,GAAG,MAAM,IAAI,CAAC,IAAI,CAAC;IACxC,MAAM,EAAE,SAAS,EAAE;IACnB,MAAM,EAAE,SAAS,EAAE;CACnB,CAAC,CAAA;AAEF,EAAE;AACF,oBAAoB;AACpB,EAAE;AAEF,MAAM,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,QAAQ,CAC7C,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,EACd,CAAC,CAAC,UAAU,CAAC,SAAS,CAAC,GAAG,CAAC,EAC3B,CAAC,CAAC,KAAK,CACN,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,EACd,CAAC,CAAC,IAAI,CAAC,aAAa,CAAC,CACrB,CACD,CAAC,CAAA;AAEF,EAAE;AACF,mBAAmB;AACnB,EAAE;AAEF,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,IAAI,CAAC,CAAC,CAAA"}

package/x/tools/common/loader.d.ts

@@ -1,19 +0,0 @@
-import { Pub } from "@e280/stz";
-import { ProgressItem } from "../speech-recognition/whisper/parts/types.js";
-export interface LoaderEvents {
-    onModelLoadProgress: Pub<ProgressItem[]>;
-    onTpsUpdate: Pub<[number]>;
-}
-export declare abstract class Loader {
-    readonly name: string;
-    model: string;
-    tps: number;
-    static loaderEvents: {
-        onModelLoadProgress: Pub<ProgressItem[]>;
-        onTpsUpdate: Pub<[number]>;
-    };
-    constructor(name: string, model: string);
-    abstract init(): Promise<void>;
-    abstract setModel(model: string): void;
-    setTps(value: number): void;
-}

package/x/tools/common/loader.js

@@ -1,18 +0,0 @@
-import { pub } from "@e280/stz";
-export class Loader {
-    name;
-    model;
-    tps = 0;
-    static loaderEvents = {
-        onModelLoadProgress: pub(),
-        onTpsUpdate: pub()
-    };
-    constructor(name, model) {
-        this.name = name;
-        this.model = model;
-    }
-    setTps(value) {
-        this.tps = value;
-    }
-}
-//# sourceMappingURL=loader.js.map

package/x/tools/common/loader.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"loader.js","sourceRoot":"","sources":["../../../s/tools/common/loader.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,GAAG,EAAM,MAAM,WAAW,CAAA;AAQlC,MAAM,OAAgB,MAAM;IAQC;IAAqB;IAPjD,GAAG,GAAG,CAAC,CAAA;IAEP,MAAM,CAAC,YAAY,GAAG;QACrB,mBAAmB,EAAE,GAAG,EAAkB;QAC1C,WAAW,EAAE,GAAG,EAAY;KAC5B,CAAA;IAED,YAA4B,IAAY,EAAS,KAAa;QAAlC,SAAI,GAAJ,IAAI,CAAQ;QAAS,UAAK,GAAL,KAAK,CAAQ;IAAG,CAAC;IAMlE,MAAM,CAAC,KAAa;QACnB,IAAI,CAAC,GAAG,GAAG,KAAK,CAAA;IACjB,CAAC"}

package/x/tools/common/transformer-pipeline.d.ts

@@ -1,8 +0,0 @@
-import { ProgressCallback } from "../speech-recognition/whisper/parts/types.js";
-export declare class PipelineFactory {
-    task: string;
-    instance: any;
-    model: string | null;
-    constructor(task: string);
-    createInstance(model: string, progressCallback?: ProgressCallback): Promise<any>;
-}

package/x/tools/common/transformer-pipeline.js

@@ -1,24 +0,0 @@
-//@ts-ignore
-import { pipeline } from "https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.7.0/dist/transformers.min.js";
-export class PipelineFactory {
-    task;
-    instance = null;
-    model = null;
-    constructor(task) {
-        this.task = task;
-    }
-    async createInstance(model, progressCallback) {
-        this.model = model;
-        return this.instance = await pipeline(this.task, this.model, {
-            dtype: {
-                encoder_model: this.model === "onnx-community/whisper-large-v3-turbo"
-                    ? "fp16"
-                    : "fp32",
-                decoder_model_merged: "q4",
-            },
-            device: "webgpu",
-            progress_callback: progressCallback,
-        });
-    }
-}
-//# sourceMappingURL=transformer-pipeline.js.map

package/x/tools/common/transformer-pipeline.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"transformer-pipeline.js","sourceRoot":"","sources":["../../../s/tools/common/transformer-pipeline.ts"],"names":[],"mappings":"AAAA,YAAY;AACZ,OAAO,EAAC,QAAQ,EAAC,MAAM,uFAAuF,CAAA;AAI9G,MAAM,OAAO,eAAe;IAIR;IAHnB,QAAQ,GAAQ,IAAI,CAAA;IACpB,KAAK,GAAkB,IAAI,CAAA;IAE3B,YAAmB,IAAY;QAAZ,SAAI,GAAJ,IAAI,CAAQ;IAAG,CAAC;IAEnC,KAAK,CAAC,cAAc,CAAC,KAAa,EAAE,gBAAmC;QACtE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAA;QAClB,OAAO,IAAI,CAAC,QAAQ,GAAG,MAAM,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,KAAK,EAAE;YAC5D,KAAK,EAAE;gBACN,aAAa,EACZ,IAAI,CAAC,KAAK,KAAK,uCAAuC;oBACrD,CAAC,CAAC,MAAM;oBACR,CAAC,CAAC,MAAM;gBACV,oBAAoB,EAAE,IAAI;aAC1B;YACD,MAAM,EAAE,QAAQ;YAChB,iBAAiB,EAAE,gBAAgB;SACnC,CAAC,CAAA;IACH,CAAC;CACD"}

package/x/tools/speech-recognition/common/model.d.ts

@@ -1,14 +0,0 @@
-import { Loader } from "../../common/loader.js";
-import { DecoderSource } from "../../../driver/fns/schematic.js";
-import { SpeechRecognizerModels, Word, WordGroup } from "../whisper/parts/types.js";
-export declare abstract class SpeechRecognizer extends Loader {
-    multilingual: boolean;
-    static speechRecognizerEvents: {
-        onTranscriptionChunk: import("@e280/stz").Pub<Word[]>;
-        onTranscribeProgress: import("@e280/stz").Pub<[number]>;
-    };
-    abstract transcribe(input: DecoderSource): Promise<WordGroup>;
-    setMultilingual(value: boolean): void;
-    detectLanguage?(input: Blob | AudioBuffer): Promise<string>;
-    setModel(value: SpeechRecognizerModels): void;
-}

package/x/tools/speech-recognition/common/model.js

@@ -1,16 +0,0 @@
-import { pub } from "@e280/stz";
-import { Loader } from "../../common/loader.js";
-export class SpeechRecognizer extends Loader {
-    multilingual = true;
-    static speechRecognizerEvents = {
-        onTranscriptionChunk: pub(),
-        onTranscribeProgress: pub()
-    };
-    setMultilingual(value) {
-        this.multilingual = value;
-    }
-    setModel(value) {
-        this.model = value;
-    }
-}
-//# sourceMappingURL=model.js.map

package/x/tools/speech-recognition/common/model.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"model.js","sourceRoot":"","sources":["../../../../s/tools/speech-recognition/common/model.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,GAAG,EAAC,MAAM,WAAW,CAAA;AAE7B,OAAO,EAAC,MAAM,EAAC,MAAM,wBAAwB,CAAA;AAI7C,MAAM,OAAgB,gBAAiB,SAAQ,MAAM;IACpD,YAAY,GAAG,IAAI,CAAA;IAEnB,MAAM,CAAC,sBAAsB,GAAG;QAC/B,oBAAoB,EAAE,GAAG,EAAU;QACnC,oBAAoB,EAAE,GAAG,EAAY;KACrC,CAAA;IAID,eAAe,CAAC,KAAc;QAC7B,IAAI,CAAC,YAAY,GAAG,KAAK,CAAA;IAC1B,CAAC;IAID,QAAQ,CAAC,KAA6B;QACrC,IAAI,CAAC,KAAK,GAAG,KAAK,CAAA;IACnB,CAAC"}

package/x/tools/speech-recognition/whisper/fns/host.d.ts

@@ -1,13 +0,0 @@
-import { ProgressItem } from "../parts/types.js";
-import { SpeechRecognizerHostEvents } from "./schematic.js";
-export declare const setupWhisperHost: (events: SpeechRecognizerHostEvents) => import("@e280/comrade").SetupHost<{
-    work: {
-        transcribe(input: import("../parts/types.js").TranscriptionMessage): Promise<import("../parts/types.js").TranscriptionResult | null>;
-    };
-    host: {
-        updateModelLoadProgress(item: ProgressItem): Promise<void>;
-        deliverTranscriptionChunk(chunk: import("../parts/types.js").TranscriptionChunk): Promise<void>;
-        updateTps(value: number): Promise<void>;
-        updateTranscribeProgress(value: number): Promise<void>;
-    };
-}>;

package/x/tools/speech-recognition/whisper/fns/host.js

@@ -1,19 +0,0 @@
-import { Comrade } from "@e280/comrade";
-export const setupWhisperHost = (events) => (Comrade.host(_shell => ({
-    async updateModelLoadProgress(item) {
-        events.onModelLoadProgress.pub(item);
-    },
-    async deliverTranscriptionChunk(chunk) {
-        events.onTranscriptionChunk.pub({
-            text: chunk.text,
-            timestamp: chunk.timestamp
-        });
-    },
-    async updateTps(value) {
-        events.onTpsUpdate.pub(value);
-    },
-    async updateTranscribeProgress(value) {
-        events.onTranscribeProgress(value);
-    }
-})));
-//# sourceMappingURL=host.js.map

package/x/tools/speech-recognition/whisper/fns/host.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"host.js","sourceRoot":"","sources":["../../../../../s/tools/speech-recognition/whisper/fns/host.ts"],"names":[],"mappings":"AACA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AAIrC,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,MAAkC,EAAE,EAAE,CAAC,CACvE,OAAO,CAAC,IAAI,CAAmB,MAAM,CAAC,EAAE,CAAC,CAAC;IACzC,KAAK,CAAC,uBAAuB,CAAC,IAAI;QACjC,MAAM,CAAC,mBAAmB,CAAC,GAAG,CAAC,IAAI,CAAC,CAAA;IACrC,CAAC;IACD,KAAK,CAAC,yBAAyB,CAAC,KAAK;QACpC,MAAM,CAAC,oBAAoB,CAAC,GAAG,CAAC;YAC/B,IAAI,EAAE,KAAK,CAAC,IAAI;YAChB,SAAS,EAAE,KAAK,CAAC,SAAS;SAC1B,CAAC,CAAA;IACH,CAAC;IACD,KAAK,CAAC,SAAS,CAAC,KAAK;QACpB,MAAM,CAAC,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC,CAAA;IAC9B,CAAC;IACD,KAAK,CAAC,wBAAwB,CAAC,KAAK;QACnC,MAAM,CAAC,oBAAoB,CAAC,KAAK,CAAC,CAAA;IACnC,CAAC;CACD,CAAC,CAAC,CACH,CAAA"}

package/x/tools/speech-recognition/whisper/fns/schematic.d.ts

@@ -1,19 +0,0 @@
-import { Pub } from "@e280/stz";
-import { AsSchematic } from "@e280/comrade";
-import { LoaderEvents } from "../../../common/loader.js";
-import { ProgressItem, TranscriptionChunk, TranscriptionMessage, TranscriptionResult, Word } from "../parts/types.js";
-export type WhisperSchematic = AsSchematic<{
-    work: {
-        transcribe(input: TranscriptionMessage): Promise<TranscriptionResult | null>;
-    };
-    host: {
-        updateModelLoadProgress(item: ProgressItem): Promise<void>;
-        deliverTranscriptionChunk(chunk: TranscriptionChunk): Promise<void>;
-        updateTps(value: number): Promise<void>;
-        updateTranscribeProgress(value: number): Promise<void>;
-    };
-}>;
-export interface SpeechRecognizerHostEvents extends LoaderEvents {
-    onTranscriptionChunk: Pub<Word[]>;
-    onTranscribeProgress: Pub<[number]>;
-}

package/x/tools/speech-recognition/whisper/fns/schematic.js

@@ -1,2 +0,0 @@
-export {};
-//# sourceMappingURL=schematic.js.map

package/x/tools/speech-recognition/whisper/fns/schematic.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"schematic.js","sourceRoot":"","sources":["../../../../../s/tools/speech-recognition/whisper/fns/schematic.ts"],"names":[],"mappings":""}

package/x/tools/speech-recognition/whisper/fns/work.d.ts

@@ -1,12 +0,0 @@
-import { TranscriptionChunk } from "../parts/types.js";
-export declare const setupWhisperWork: import("@e280/comrade").SetupWork<{
-    work: {
-        transcribe(input: import("../parts/types.js").TranscriptionMessage): Promise<import("../parts/types.js").TranscriptionResult | null>;
-    };
-    host: {
-        updateModelLoadProgress(item: import("../parts/types.js").ProgressItem): Promise<void>;
-        deliverTranscriptionChunk(chunk: TranscriptionChunk): Promise<void>;
-        updateTps(value: number): Promise<void>;
-        updateTranscribeProgress(value: number): Promise<void>;
-    };
-}>;

package/x/tools/speech-recognition/whisper/fns/work.js

@@ -1,74 +0,0 @@
-import { Comrade } from "@e280/comrade";
-//@ts-ignore
-import { WhisperTextStreamer } from "https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.7.0/dist/transformers.min.js";
-import { PipelineFactory } from "../../../common/transformer-pipeline.js";
-// TODO suspicious globals, probably bad
-const pipeline = new PipelineFactory("automatic-speech-recognition");
-let transcriber;
-export const setupWhisperWork = Comrade.work(shell => ({
-    async transcribe({ audio, model, language, duration }) {
-        const isDistil = model.startsWith("distil-whisper/");
-        if (!pipeline.model || pipeline.model !== model) {
-            pipeline.instance?.dispose()?.();
-            pipeline.instance = null;
-            transcriber = await pipeline.createInstance(model, (data) => {
-                if (data.progress)
-                    shell.host.updateModelLoadProgress({
-                        id: data.file,
-                        progress: data.progress
-                    });
-            });
-        }
-        const timePrecision = transcriber.processor.feature_extractor.config.chunk_length /
-            transcriber.model.config.max_source_positions;
-        const chunkLength = isDistil ? 20 : 30;
-        const strideLength = isDistil ? 3 : 5;
-        let chunkCount = 0;
-        let startTime = null;
-        let tokenCount = 0;
-        let tps = 0;
-        const chunkDuration = chunkLength - strideLength;
-        const estimateProgress = () => {
-            const audioProgressSeconds = chunkCount * chunkDuration;
-            return Math.min(audioProgressSeconds / duration, 1);
-        };
-        const streamer = new WhisperTextStreamer(transcriber.tokenizer, {
-            time_precision: timePrecision,
-            token_callback_function: () => {
-                startTime ??= performance.now();
-                if (++tokenCount > 1) {
-                    tps = (tokenCount / (performance.now() - startTime)) * 1000;
-                    shell.host.updateTps(tps);
-                }
-            },
-            callback_function: (textChunk) => {
-                shell.host.deliverTranscriptionChunk(textChunk);
-            },
-            on_finalize: () => {
-                startTime = null;
-                tokenCount = 0;
-                chunkCount++;
-                const progress = estimateProgress();
-                shell.host.updateTranscribeProgress(progress);
-            },
-        });
-        const output = await transcriber(audio, {
-            top_k: 0,
-            do_sample: false,
-            chunk_length_s: chunkLength,
-            stride_length_s: strideLength,
-            language,
-            task: "transcribe",
-            return_timestamps: "word", // if using "word" the on_chunk_start & end is not called thus we cant retrieve timestamps, only after whole thing finishes
-            force_full_sequences: false,
-            streamer,
-        });
-        if (!output)
-            return null;
-        return {
-            tps,
-            ...output,
-        };
-    }
-}));
-//# sourceMappingURL=work.js.map

package/x/tools/speech-recognition/whisper/fns/work.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"work.js","sourceRoot":"","sources":["../../../../../s/tools/speech-recognition/whisper/fns/work.ts"],"names":[],"mappings":"AAAA,OAAO,EAAC,OAAO,EAAC,MAAM,eAAe,CAAA;AACrC,YAAY;AACZ,OAAO,EAAW,mBAAmB,EAAC,MAAM,uFAAuF,CAAA;AAInI,OAAO,EAAC,eAAe,EAAC,MAAM,yCAAyC,CAAA;AAEvE,wCAAwC;AACxC,MAAM,QAAQ,GAAG,IAAI,eAAe,CAAC,8BAA8B,CAAC,CAAA;AACpE,IAAI,WAAgB,CAAA;AAEpB,MAAM,CAAC,MAAM,gBAAgB,GAAG,OAAO,CAAC,IAAI,CAAmB,KAAK,CAAC,EAAE,CAAC,CAAC;IACxE,KAAK,CAAC,UAAU,CAAC,EAAC,KAAK,EAAE,KAAK,EAAE,QAAQ,EAAE,QAAQ,EAAC;QAClD,MAAM,QAAQ,GAAG,KAAK,CAAC,UAAU,CAAC,iBAAiB,CAAC,CAAA;QAEpD,IAAG,CAAC,QAAQ,CAAC,KAAK,IAAI,QAAQ,CAAC,KAAK,KAAK,KAAK,EAAE,CAAC;YAChD,QAAQ,CAAC,QAAQ,EAAE,OAAO,EAAE,EAAE,EAAE,CAAA;YAChC,QAAQ,CAAC,QAAQ,GAAG,IAAI,CAAA;YACxB,WAAW,GAAG,MAAM,QAAQ,CAAC,cAAc,CAC1C,KAAK,EACL,CAAC,IAAI,EAAE,EAAE;gBACR,IAAG,IAAI,CAAC,QAAQ;oBACf,KAAK,CAAC,IAAI,CAAC,uBAAuB,CAAC;wBAClC,EAAE,EAAE,IAAI,CAAC,IAAI;wBACb,QAAQ,EAAE,IAAI,CAAC,QAAQ;qBACvB,CAAC,CAAA;YACJ,CAAC,CACD,CAAA;QACF,CAAC;QAED,MAAM,aAAa,GAClB,WAAW,CAAC,SAAS,CAAC,iBAAiB,CAAC,MAAM,CAAC,YAAY;YAC3D,WAAW,CAAC,KAAK,CAAC,MAAM,CAAC,oBAAoB,CAAA;QAE9C,MAAM,WAAW,GAAG,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAA;QACtC,MAAM,YAAY,GAAG,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAA;QAErC,IAAI,UAAU,GAAG,CAAC,CAAA;QAClB,IAAI,SAAS,GAAkB,IAAI,CAAA;QACnC,IAAI,UAAU,GAAG,CAAC,CAAA;QAClB,IAAI,GAAG,GAAG,CAAC,CAAA;QAEX,MAAM,aAAa,GAAG,WAAW,GAAG,YAAY,CAAA;QAEhD,MAAM,gBAAgB,GAAG,GAAG,EAAE;YAC7B,MAAM,oBAAoB,GAAG,UAAU,GAAG,aAAa,CAAA;YACvD,OAAO,IAAI,CAAC,GAAG,CAAC,oBAAoB,GAAG,QAAQ,EAAE,CAAC,CAAC,CAAA;QACpD,CAAC,CAAA;QAED,MAAM,QAAQ,GAAG,IAAI,mBAAmB,CAAC,WAAW,CAAC,SAAS,EAAE;YAC/D,cAAc,EAAE,aAAa;YAC7B,uBAAuB,EAAE,GAAG,EAAE;gBAC7B,SAAS,KAAK,WAAW,CAAC,GAAG,EAAE,CAAA;gBAC/B,IAAI,EAAE,UAAU,GAAG,CAAC,EAAE,CAAC;oBACtB,GAAG,GAAG,CAAC,UAAU,GAAG,CAAC,WAAW,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC,CAAC,GAAG,IAAI,CAAA;oBAC3D,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,CAAA;gBAC1B,CAAC;YACF,CAAC;YACD,iBAAiB,EAAE,CAAC,SAAc,EAAE,EAAE;gBACrC,KAAK,CAAC,IAAI,CAAC,yBAAyB,CAAC,SAAS,CAAC,CAAA;YAChD,CAAC;YACD,WAAW,EAAE,GAAG,EAAE;gBACjB,SAAS,GAAG,IAAI,CAAA;gBAChB,UAAU,GAAG,CAAC,CAAA;gBACd,UAAU,EAAE,CAAA;gBACZ,MAAM,QAAQ,GAAG,gBAAgB,EAAE,CAAA;gBACnC,KAAK,CAAC,IAAI,CAAC,wBAAwB,CAAC,QAAQ,CAAC,CAAA;YAC9C,CAAC;SACD,CAAC,CAAA;QAEF,MAAM,MAAM,GAAG,MAAM,WAAW,CAAC,KAAK,EAAE;YACvC,KAAK,EAAE,CAAC;YACR,SAAS,EAAE,KAAK;YAChB,cAAc,EAAE,WAAW;YAC3B,eAAe,EAAE,YAAY;YAC7B,QAAQ;YACR,IAAI,EAAE,YAAY;YAClB,iBAAiB,EAAE,MAAM,EAAE,2HAA2H;YACtJ,oBAAoB,EAAE,KAAK;YAC3B,QAAQ;SACR,CAAC,CAAA;QAEF,IAAI,CAAC,MAAM;YAAE,OAAO,IAAI,CAAA;QAExB,OAAO;YACN,GAAG;YACH,GAAG,MAAM;SACT,CAAA;IACF,CAAC;CACD,CAAC,CAAC,CAAA"}

package/x/tools/speech-recognition/whisper/parts/types.d.ts

@@ -1,31 +0,0 @@
-export interface ProgressItem {
-    id: string;
-    progress: number;
-}
-export type Word = {
-    text: string;
-    timestamp: [start: number, end: number];
-};
-export type WordGroup = Word[];
-export type Transcript = WordGroup[];
-export interface TranscriptionChunk {
-    text: string;
-    offset: number;
-    timestamp: [number, number];
-    finalised: boolean;
-}
-export interface TranscriptionMessage {
-    audio: Float32Array;
-    model: string;
-    subtask: string | null;
-    language: string | null;
-    duration: number;
-}
-export interface TranscriptionResult {
-    text: string;
-    chunks: TranscriptionChunk[];
-    tps: number;
-}
-export type ProgressCallback = (data: any) => void;
-export type SpeechRecognizerModels = "onnx-community/whisper-tiny_timestamped";
-export type SpeechRecognizerSubtasks = "transcribe";

package/x/tools/speech-recognition/whisper/parts/types.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"types.js","sourceRoot":"","sources":["../../../../../s/tools/speech-recognition/whisper/parts/types.ts"],"names":[],"mappings":""}

package/x/tools/speech-recognition/whisper/parts/worker.bundle.d.ts

@@ -1 +0,0 @@
-export {};

package/x/tools/speech-recognition/whisper/parts/worker.bundle.js

@@ -1,4 +0,0 @@
-import { Comrade } from "@e280/comrade";
-import { setupWhisperWork } from "../fns/work.js";
-await Comrade.worker(setupWhisperWork);
-//# sourceMappingURL=worker.bundle.js.map