@remotion/openai-whisper 4.0.278 → 4.0.280

package/dist/openai-whisper-api-to-captions.js

@@ -10,7 +10,13 @@ const openAiWhisperApiToCaptions = ({ transcription, }) => {
         throw new Error('The transcription does need to be been generated with `timestamp_granularities: ["word"]`');
     }
     let remainingText = transcription.text;
-    for (const word of transcription.words) {
+    for (let i = 0; i < transcription.words.length; i++) {
+        const word = transcription.words[i];
+        const firstWord = i === 0;
+        // https://github.com/remotion-dev/remotion/issues/5031
+        if (firstWord) {
+            word.word = word.word.trimStart();
+        }
         const punctuation = `\\?,\\.\\%\\–\\!\\;\\:\\'\\"\\-\\_\\(\\)\\[\\]\\{\\}\\@\\#\\$\\^\\&\\*\\+\\=\\/\\|\\<\\>\\~\``;
         const match = new RegExp(`^([\\s${punctuation}]{0,4})${word.word.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')}([${punctuation}]{0,3})?`).exec(remainingText);
         if (!match) {

package/dist/test/foreign-api.test.d.ts

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

package/dist/test/foreign-api.test.js

@@ -0,0 +1,278 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+const bun_test_1 = require("bun:test");
+const openai_whisper_api_to_captions_1 = require("../openai-whisper-api-to-captions");
+(0, bun_test_1.test)('foreign API (https://github.com/remotion-dev/remotion/issues/5031)', () => {
+    const res = {
+        text: "Hello world. It's great to finally be here, surrounded by the endless possibilities that this universe has to offer, and I'm excited to see what the future holds for all of us.",
+        task: 'transcribe',
+        language: '',
+        duration: 10.625,
+        words: [
+            { word: ' Hello', start: 0, end: 0.46000000834465027 },
+            { word: ' world.', start: 0.46000000834465027, end: 0.9800000190734863 },
+            { word: " It's", start: 0.9800000190734863, end: 1.5800000429153442 },
+            { word: ' great', start: 1.5800000429153442, end: 1.7999999523162842 },
+            { word: ' to', start: 1.7999999523162842, end: 1.9600000381469727 },
+            { word: ' finally', start: 1.9600000381469727, end: 2.319999933242798 },
+            { word: ' be', start: 2.319999933242798, end: 2.5799999237060547 },
+            { word: ' here,', start: 2.5799999237060547, end: 3 },
+            { word: ' surrounded', start: 3, end: 3.5999999046325684 },
+            { word: ' by', start: 3.5999999046325684, end: 4.019999980926514 },
+            { word: ' the', start: 4.019999980926514, end: 4.179999828338623 },
+            { word: ' endless', start: 4.179999828338623, end: 4.519999980926514 },
+            {
+                word: ' possibilities',
+                start: 4.519999980926514,
+                end: 5.119999885559082,
+            },
+            { word: ' that', start: 5.119999885559082, end: 5.639999866485596 },
+            { word: ' this', start: 5.639999866485596, end: 5.820000171661377 },
+            { word: ' universe', start: 5.820000171661377, end: 6.179999828338623 },
+            { word: ' has', start: 6.179999828338623, end: 6.519999980926514 },
+            { word: ' to', start: 6.519999980926514, end: 6.659999847412109 },
+            { word: ' offer,', start: 6.659999847412109, end: 7.179999828338623 },
+            { word: ' and', start: 7.179999828338623, end: 7.679999828338623 },
+            { word: " I'm", start: 7.679999828338623, end: 7.820000171661377 },
+            { word: ' excited', start: 7.820000171661377, end: 8.180000305175781 },
+            { word: ' to', start: 8.180000305175781, end: 8.420000076293945 },
+            { word: ' see', start: 8.420000076293945, end: 8.600000381469727 },
+            { word: ' what', start: 8.600000381469727, end: 8.739999771118164 },
+            { word: ' the', start: 8.739999771118164, end: 8.859999656677246 },
+            { word: ' future', start: 8.859999656677246, end: 9.180000305175781 },
+            { word: ' holds', start: 9.180000305175781, end: 9.520000457763672 },
+            { word: ' for', start: 9.520000457763672, end: 9.779999732971191 },
+            { word: ' all', start: 9.779999732971191, end: 9.979999542236328 },
+            { word: ' of', start: 9.979999542236328, end: 10.079999923706055 },
+            { word: ' us.', start: 10.079999923706055, end: 10.34000015258789 },
+        ],
+    };
+    (0, bun_test_1.expect)((0, openai_whisper_api_to_captions_1.openAiWhisperApiToCaptions)({ transcription: res })).toEqual({
+        captions: [
+            {
+                confidence: null,
+                endMs: 460.00000834465027,
+                startMs: 0,
+                text: 'Hello',
+                timestampMs: 230.00000417232513,
+            },
+            {
+                confidence: null,
+                endMs: 980.0000190734863,
+                startMs: 460.00000834465027,
+                text: ' world.',
+                timestampMs: 720.0000137090683,
+            },
+            {
+                confidence: null,
+                endMs: 1580.0000429153442,
+                startMs: 980.0000190734863,
+                text: " It's",
+                timestampMs: 1280.0000309944153,
+            },
+            {
+                confidence: null,
+                endMs: 1799.9999523162842,
+                startMs: 1580.0000429153442,
+                text: ' great',
+                timestampMs: 1689.9999976158142,
+            },
+            {
+                confidence: null,
+                endMs: 1960.0000381469727,
+                startMs: 1799.9999523162842,
+                text: ' to',
+                timestampMs: 1879.9999952316284,
+            },
+            {
+                confidence: null,
+                endMs: 2319.999933242798,
+                startMs: 1960.0000381469727,
+                text: ' finally',
+                timestampMs: 2139.9999856948853,
+            },
+            {
+                confidence: null,
+                endMs: 2579.9999237060547,
+                startMs: 2319.999933242798,
+                text: ' be',
+                timestampMs: 2449.9999284744263,
+            },
+            {
+                confidence: null,
+                endMs: 3000,
+                startMs: 2579.9999237060547,
+                text: ' here,',
+                timestampMs: 2789.9999618530273,
+            },
+            {
+                confidence: null,
+                endMs: 3599.9999046325684,
+                startMs: 3000,
+                text: ' surrounded',
+                timestampMs: 3299.999952316284,
+            },
+            {
+                confidence: null,
+                endMs: 4019.9999809265137,
+                startMs: 3599.9999046325684,
+                text: ' by',
+                timestampMs: 3809.999942779541,
+            },
+            {
+                confidence: null,
+                endMs: 4179.999828338623,
+                startMs: 4019.9999809265137,
+                text: ' the',
+                timestampMs: 4099.999904632568,
+            },
+            {
+                confidence: null,
+                endMs: 4519.999980926514,
+                startMs: 4179.999828338623,
+                text: ' endless',
+                timestampMs: 4349.999904632568,
+            },
+            {
+                confidence: null,
+                endMs: 5119.999885559082,
+                startMs: 4519.999980926514,
+                text: ' possibilities',
+                timestampMs: 4819.999933242798,
+            },
+            {
+                confidence: null,
+                endMs: 5639.999866485596,
+                startMs: 5119.999885559082,
+                text: ' that',
+                timestampMs: 5379.999876022339,
+            },
+            {
+                confidence: null,
+                endMs: 5820.000171661377,
+                startMs: 5639.999866485596,
+                text: ' this',
+                timestampMs: 5730.000019073486,
+            },
+            {
+                confidence: null,
+                endMs: 6179.999828338623,
+                startMs: 5820.000171661377,
+                text: ' universe',
+                timestampMs: 6000,
+            },
+            {
+                confidence: null,
+                endMs: 6519.999980926514,
+                startMs: 6179.999828338623,
+                text: ' has',
+                timestampMs: 6349.999904632568,
+            },
+            {
+                confidence: null,
+                endMs: 6659.999847412109,
+                startMs: 6519.999980926514,
+                text: ' to',
+                timestampMs: 6589.9999141693115,
+            },
+            {
+                confidence: null,
+                endMs: 7179.999828338623,
+                startMs: 6659.999847412109,
+                text: ' offer,',
+                timestampMs: 6919.999837875366,
+            },
+            {
+                confidence: null,
+                endMs: 7679.999828338623,
+                startMs: 7179.999828338623,
+                text: ' and',
+                timestampMs: 7429.999828338623,
+            },
+            {
+                confidence: null,
+                endMs: 7820.000171661377,
+                startMs: 7679.999828338623,
+                text: " I'm",
+                timestampMs: 7750,
+            },
+            {
+                confidence: null,
+                endMs: 8180.000305175781,
+                startMs: 7820.000171661377,
+                text: ' excited',
+                timestampMs: 8000.000238418579,
+            },
+            {
+                confidence: null,
+                endMs: 8420.000076293945,
+                startMs: 8180.000305175781,
+                text: ' to',
+                timestampMs: 8300.000190734863,
+            },
+            {
+                confidence: null,
+                endMs: 8600.000381469727,
+                startMs: 8420.000076293945,
+                text: ' see',
+                timestampMs: 8510.000228881836,
+            },
+            {
+                confidence: null,
+                endMs: 8739.999771118164,
+                startMs: 8600.000381469727,
+                text: ' what',
+                timestampMs: 8670.000076293945,
+            },
+            {
+                confidence: null,
+                endMs: 8859.999656677246,
+                startMs: 8739.999771118164,
+                text: ' the',
+                timestampMs: 8799.999713897705,
+            },
+            {
+                confidence: null,
+                endMs: 9180.000305175781,
+                startMs: 8859.999656677246,
+                text: ' future',
+                timestampMs: 9019.999980926514,
+            },
+            {
+                confidence: null,
+                endMs: 9520.000457763672,
+                startMs: 9180.000305175781,
+                text: ' holds',
+                timestampMs: 9350.000381469727,
+            },
+            {
+                confidence: null,
+                endMs: 9779.999732971191,
+                startMs: 9520.000457763672,
+                text: ' for',
+                timestampMs: 9650.000095367432,
+            },
+            {
+                confidence: null,
+                endMs: 9979.999542236328,
+                startMs: 9779.999732971191,
+                text: ' all',
+                timestampMs: 9879.99963760376,
+            },
+            {
+                confidence: null,
+                endMs: 10079.999923706055,
+                startMs: 9979.999542236328,
+                text: ' of',
+                timestampMs: 10029.999732971191,
+            },
+            {
+                confidence: null,
+                endMs: 10340.00015258789,
+                startMs: 10079.999923706055,
+                text: ' us.',
+                timestampMs: 10210.000038146973,
+            },
+        ],
+    });
+});

package/package.json

@@ -3,7 +3,7 @@
     "url": "https://github.com/remotion-dev/remotion/tree/main/packages/openai-whisper"
   },
   "name": "@remotion/openai-whisper",
-  "version": "4.0.278",
+  "version": "4.0.280",
   "description": "Work with the output of the OpenAI Whisper API",
   "main": "dist/index.js",
   "sideEffects": false,
@@ -16,13 +16,13 @@
   "author": "Jonny Burger <jonny@remotion.dev>",
   "license": "MIT",
   "dependencies": {
-    "@remotion/captions": "4.0.278"
+    "@remotion/captions": "4.0.280"
   },
   "peerDependencies": {},
   "devDependencies": {
     "openai": "4.67.1",
     "eslint": "9.19.0",
-    "@remotion/eslint-config-internal": "4.0.278"
+    "@remotion/eslint-config-internal": "4.0.280"
   },
   "keywords": [
     "remotion"