{
  "id": 14604818,
  "title": "Why live dubbing lags a sentence behind, and how dubbing from buffered audio fixes it",
  "url": "https://urgent.news/2026/10/15/why-live-dubbing-lags-a-sentence-behind-and-how-dubbing-from-buffered",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-15T03:46:53.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/chriscornell/why-live-dubbing-lags-a-sentence-behind-and-how-dubbing-from-buffered-audio-fixes-it-5pe"
  },
  "original_language": "en",
  "account": "When watching videos on platforms like YouTube or Bilibili, you may have noticed the dubbed audio lags a sentence behind the speaker. This delay occurs because real-time video translators have to wait for the speaker to finish a sentence before translating it. Translating a complete sentence is crucial for accuracy, as word order can vary significantly between languages. The translation process then involves running speech recognition, translating the sentence, synthesizing the translated speech, and finally playing it. These steps take time, contributing to the lag.\n\nHowever, when it comes to recorded videos, the situation is different. The browser's media player pre-downloads upcoming audio, which means the audio data is already available before the viewer hears it. This allows the system to start translation ahead of the viewer, effectively removing the one-sentence lag. The pipeline for this process involves recognizing speech from the buffered audio, joining fragments into complete sentences, translating the sentences, and synchronizing the synthesized speech with the original timing of the speaker.\n\nPractical considerations are important when implementing live dubbing. The system doesn't need to process the entire video upfront, as only a small window of buffered audio ahead of the playhead needs to be ready. If the video is paused or the user jumps to a new point, the system must refill the window before delivering the next translated sentence. Additionally, translated sentences might not always fit within the audio, so the system can gently speed up the translation or skip a sentence to maintain quality. Using context from the video's title and surrounding sentences helps maintain consistency for technical terms and names throughout a tutorial.\n\nHowever, live dubbing has its limitations. It only works with recorded videos, not live streams. Loud music, overlapping speakers, and heavy accents can still lead to recognition errors. Additionally, the system generates a single voice for all speakers, as recognized speech does not include speaker identification. In cases where videos already have captions, including auto-generated ones, the system skips the speech recognition step and relies on the caption timings to sync the translation. This highlights the adaptability of the approach, as it can function effectively with either audio-only or captioned videos.",
  "summary": "Disclosure: I work on EarDub, a browser extension that dubs videos into other languages. This post is about a design problem we hit, not a product pitch, but the product comes up at the end. If you've tried a \"real-time\" video translator that listens to a video and speaks a translation, you've probably noticed the voice runs about one sentence behind the speaker. The person on screen finishes a…",
  "key_points": [
    "Live dubbing lags a sentence behind due to translation process.",
    "Buffered audio allows translation to start ahead of viewer.",
    "Works with recorded videos, not live streams."
  ],
  "editors_take": "Leveraging buffered audio for recorded content allows translation systems to bypass the inherent delays of real-time speech processing, providing a more synchronized experience than live stream alternatives.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}