@remotion/install-whisper-cpp 4.0.129 → 4.0.131

package/dist/transcribe.d.ts

@@ -7,11 +7,22 @@ type Offsets = {
     from: number;
     to: number;
 };
+type WordLevelToken = {
+    t_dtw: number;
+    text: string;
+    timestamps: Timestamps;
+    offsets: Offsets;
+    id: number;
+    p: number;
+};
 type TranscriptionItem = {
     timestamps: Timestamps;
     offsets: Offsets;
     text: string;
 };
+type TranscriptionItemWithTimestamp = TranscriptionItem & {
+    tokens: WordLevelToken[];
+};
 type Model = {
     type: string;
     multilingual: boolean;
@@ -39,18 +50,19 @@ type Params = {
 type Result = {
     language: string;
 };
-export type TranscriptionJson = {
+export type TranscriptionJson<WithTokenLevelTimestamp extends boolean> = {
     systeminfo: string;
     model: Model;
     params: Params;
     result: Result;
-    transcription: TranscriptionItem[];
+    transcription: true extends WithTokenLevelTimestamp ? TranscriptionItemWithTimestamp[] : TranscriptionItem[];
 };
-export declare const transcribe: ({ inputPath, whisperPath, model, modelFolder, translateToEnglish, }: {
+export declare const transcribe: <HasTokenLevelTimestamps extends boolean>({ inputPath, whisperPath, model, modelFolder, translateToEnglish, tokenLevelTimestamps, }: {
     inputPath: string;
     whisperPath: string;
     model: WhisperModel;
+    tokenLevelTimestamps: HasTokenLevelTimestamps;
     modelFolder?: string | undefined;
     translateToEnglish?: boolean | undefined;
-}) => Promise<TranscriptionJson>;
+}) => Promise<TranscriptionJson<HasTokenLevelTimestamps>>;
 export {};

package/dist/transcribe.js

@@ -31,7 +31,6 @@ const node_child_process_1 = require("node:child_process");
 const node_fs_1 = __importStar(require("node:fs"));
 const node_os_1 = __importDefault(require("node:os"));
 const node_path_1 = __importDefault(require("node:path"));
-const node_util_1 = __importDefault(require("node:util"));
 const download_whisper_model_1 = require("./download-whisper-model");
 const isWavFile = (inputPath) => {
     const splitted = inputPath.split('.');
@@ -42,27 +41,58 @@ const isWavFile = (inputPath) => {
 };
 const readJson = async (jsonPath) => {
     const data = await node_fs_1.default.promises.readFile(jsonPath, 'utf8');
-    const jsonData = await JSON.parse(data);
-    return jsonData;
+    return JSON.parse(data);
 };
-const transcribeToTempJSON = async ({ fileToTranscribe, whisperPath, model, tmpJSONPath, modelFolder, translate, }) => {
-    const promisifiedExec = node_util_1.default.promisify(node_child_process_1.exec);
+const transcribeToTempJSON = async ({ fileToTranscribe, whisperPath, model, tmpJSONPath, modelFolder, translate, tokenLevelTimestamps, }) => {
     const modelPath = (0, download_whisper_model_1.getModelPath)(modelFolder !== null && modelFolder !== void 0 ? modelFolder : whisperPath, model);
     if (!node_fs_1.default.existsSync(modelPath)) {
-        throw new Error(`Error: Model ${model} does not exist at ${modelFolder ? modelFolder : modelPath}. Check out the downloadWhisperMode() API at https://www.remotion.dev/docs/install-whisper-cpp/download-whisper-model to see how to install whisper models`);
+        throw new Error(`Error: Model ${model} does not exist at ${modelFolder ? modelFolder : modelPath}. Check out the downloadWhisperModel() API at https://www.remotion.dev/docs/install-whisper-cpp/download-whisper-model to see how to install whisper models`);
     }
     const executable = node_os_1.default.platform() === 'win32'
         ? node_path_1.default.join(whisperPath, 'main.exe')
         : node_path_1.default.join(whisperPath, './main');
-    const modelOption = model ? `-m ${modelPath}` : '';
-    const translateOption = translate ? `-tr ` : '';
-    await promisifiedExec(`${executable} -f ${fileToTranscribe} --output-file ${tmpJSONPath} --output-json --max-len 1 ${modelOption} ${translateOption}`, { cwd: whisperPath }).then(({ stderr }) => {
-        if (stderr.includes('error')) {
-            throw new Error('An error occured while transcribing: ' + stderr);
-        }
+    const args = [
+        '-f',
+        fileToTranscribe,
+        '--output-file',
+        tmpJSONPath,
+        '--output-json',
+        '--max-len',
+        '1',
+        '-ofj',
+        tokenLevelTimestamps ? ['--dtw', model] : null,
+        model ? [`-m`, `${modelPath}`] : null,
+        translate ? '-tr' : null,
+    ]
+        .flat(1)
+        .filter(Boolean);
+    const outputPath = await new Promise((resolve, reject) => {
+        const task = (0, node_child_process_1.spawn)(executable, args, { cwd: whisperPath });
+        const predictedPath = `${tmpJSONPath}.json`;
+        let output = '';
+        const onData = (data) => {
+            const str = data.toString('utf-8');
+            output += str;
+            // Sometimes it hangs here
+            if (str.includes('ggml_metal_free: deallocating')) {
+                task.kill();
+            }
+        };
+        task.stdout.on('data', onData);
+        task.stderr.on('data', onData);
+        task.on('exit', (code) => {
+            // Whisper sometimes files also with error code 0
+            // https://github.com/ggerganov/whisper.cpp/pull/1952/files
+            if ((0, node_fs_1.existsSync)(predictedPath)) {
+                resolve(predictedPath);
+                return;
+            }
+            reject(new Error(`No transcription was created (process exited with code ${code}): ${output}`));
+        });
     });
+    return { outputPath };
 };
-const transcribe = async ({ inputPath, whisperPath, model, modelFolder, translateToEnglish = false, }) => {
+const transcribe = async ({ inputPath, whisperPath, model, modelFolder, translateToEnglish = false, tokenLevelTimestamps, }) => {
     if (!(0, node_fs_1.existsSync)(whisperPath)) {
         throw new Error(`Whisper does not exist at ${whisperPath}. Double-check the passed whisperPath. If you havent installed whisper, check out the installWhisperCpp() API at https://www.remotion.dev/docs/install-whisper-cpp/install-whisper-cpp to see how to install whisper programatically.`);
     }
@@ -73,15 +103,15 @@ const transcribe = async ({ inputPath, whisperPath, model, modelFolder, translat
         throw new Error('Invalid inputFile type. The provided file is not a wav file!');
     }
     const tmpJSONDir = node_path_1.default.join(process.cwd(), 'tmp');
-    await transcribeToTempJSON({
+    const { outputPath: tmpJSONPath } = await transcribeToTempJSON({
         fileToTranscribe: inputPath,
         whisperPath,
         model,
         tmpJSONPath: tmpJSONDir,
         modelFolder: modelFolder !== null && modelFolder !== void 0 ? modelFolder : null,
         translate: translateToEnglish,
+        tokenLevelTimestamps,
     });
-    const tmpJSONPath = `${tmpJSONDir}.json`;
     const json = (await readJson(tmpJSONPath));
     node_fs_1.default.unlinkSync(tmpJSONPath);
     return json;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@remotion/install-whisper-cpp",
-  "version": "4.0.129",
+  "version": "4.0.131",
   "description": "Install helper for Whisper.cpp",
   "main": "dist/index.js",
   "sideEffects": false,
@@ -43,6 +43,7 @@
     "formatting": "prettier src --check",
     "lint": "eslint src --ext ts,tsx",
     "build": "tsc -d",
-    "watch": "tsc -w"
+    "watch": "tsc -w",
+    "test": "vitest --run"
   }
 }