工作清單

Robots in China gear up for 2nd annual World Humanoid Games

逐字稿文字預覽

They can perform dance moves and showoff roundhouse kicks or maybe offer aride.>> Hi.>> Robots, it seems, are everywhere in
China in all shapes, sizes, and speedswith companies rolling out more powerfulhumanoid resources.This robot conference is show of intense
competition among startups here.And happening alongside it, the secondannual humanoid games in Beijing.
Where nearly 700 teams are expected tocompete for gold, more than double thesize of last year's. It's all a sign of
how seriously China is betting onrobotics as the next industrialrevolution. These days, there's a lot ofhype around this company, Unitree.
That's one of the world's biggest makersof humanoids and robot dogs. The companyhas just started trading publicly on the
Shanghai Stock Exchange and sharespopped 500%on day one.Its latest model may explain why. Ahumanoid called Superman that can jump 6
and 1/2 feet and claims to sprint fasterthan world record holder Usain Bolt,although it crashed into a wall right
after crossing the finish line.Still, it's leaps and bounds ahead ofwhere sporting humanoids were last year
when we attended China's first everobot games.>> This event is being billed as the firstof its kind.>> Teams competing in soccer, track events,
and tasks like moving luggage.>> There are hundreds of robot companieshere in China making machines that can
do backflips or run faster than peopleif that's what you want your robot todo. But a big focus here at the World
Robot Conference in Beijing is on robotsthat can do tasks that are, well,mundane or boring even by humanstandards.
>> Daily grind jobs like folding laundry,picking up toys, or sorting packages ina warehouse. Practical work that humans
might not mind giving up. Or at least,that's the way robot developers here seethe future.>> So, we don't want to replace the first
person. We we want to be people to tohave better job, to have a better life.>> The challenge here is turningeye-catching kicks and moves into actual
productivity, then scaling itcommercially. Low-cost manufacturinghere has allowed China's industry to
pull ahead. Hefty government subsidieshave helped out, too.Robots are not without some glitchesthat come with real questions about
safety. Remember when this one kicked achild?
This year, there are no robots made inthe United States, and none of what's ondisplay here in China will be finding
its way to the US anytime soon. TheTrump administration banning imports ofnew foreign-made quadruped and humanoid
robots, citing national security andprivacy concerns. A move that heavilytargets China, which, according to
consulting firm IDC, accounts for 82% ofthe world's humanoid shipments.>> There are risks to the approach that the
US is taking, but I think their heartsare in the right place. They want tobuild a supply chain. They want toreshore manufacturing. And most
importantly, they are worried aboutnational security risks.>> After the recent ban, IDC estimateshumanoid sales by US companies could
rop 58% by 2030 under a worst-casescenario. We reached out to the WhiteHouse for comment and haven't heardback. FCC Chair Brendan Carr telling NBC
News when the regulation was announced.>> So, one of the upsides of these actionis not just national security,
but it is helping to re-industrializethis country.>> China's leadership has made robotics anational priority. As the US moves to
restrict technology, China is alreadyputting the machines to work. JaniceMackey Frayer, NBC News, Beijing.
>> We thank you for watching and rememberstay updated on breaking news and topstories on the NBC News app or watch
live on our YouTube channel.

來源：transcript.txt

本次執行流程與問題處理

  1. 下達指令

    指令：videonote https://youtu.be/V9z-kLwst90?si=GrXOptZ5oP7hrwAJ --language en --bilingual --learning --auto-source-subtitles

    來源：https://youtu.be/V9z-kLwst90?si=GrXOptZ5oP7hrwAJ

    辨識語言：en ｜ 參數：--language en --bilingual --learning --auto-source-subtitles

  2. 預期流程

    依指令產生的處理順序：

    1. 自動字幕優先，Whisper 作為備援
    2. 來源選擇與品質檢查
    3. 文字校正與句子單元
    4. 雙語翻譯
    5. 雙語時間軸與學習單元
    6. 筆記與發布
    7. 公開網頁發布

  3. 1. 下載與語音辨識

    工具：videonote-transcribe-url（主辨識管線：下載、Whisper 辨識與選稿）

    處理：下載音訊、以 Whisper 辨識並選出最佳逐字稿

    產出：yt_dlp_meta.json（影片元資料）、audio_16k_mono.wav（16kHz 單聲道音訊）、transcript.whisper.json（Whisper 辨識結果）、transcript.json（正式逐字稿）、transcript_selection_report.json（來源選擇報告）

    耗時：3.5 秒

    關係：依序執行

    結果：完成

  4. 2. 全文快速校讀

    工具：videonote_rapid_transcript_review.py

    處理：先速讀完整逐字稿，再修正高可信突兀詞句並列出待確認項目

    產出：rapid_transcript_review_report.json（產出檔案）

    耗時：37.5 秒

    關係：依序執行

    結果：完成

  5. 3. 建立句子單元

    工具：videonote-build-sentence-units

    處理：將逐字稿切成帶時間軸的句子單元

    產出：sentence_units.json（句子單元）

    耗時：0.6 秒

    關係：依序執行

    結果：完成

  6. 4. 時間軸稽核

    工具：videonote_timeline_audit.py

    處理：檢查時間軸與音訊對齊

    產出：timeline_audit_report.json（時間軸稽核）

    耗時：0.6 秒

    關係：依序執行

    結果：完成

  7. 5. 品質檢查

    工具：videonote_controlled_quality.py（檢查時間軸與品質）

    處理：檢查語言、時間軸與可讀性品質

    產出：controlled_quality_report.json（品質報告）、language_sanity_report.json（語言檢查）

    耗時：0.7 秒

    關係：依序執行

    結果：完成

  8. 6. 播放時間軸回填

    工具：videonote_rebase_display_timeline.py

    處理：將網頁播放時間軸回填為校正字幕時間

    產出：display_timeline_rebase_report.json（時間軸回填報告）

    耗時：0.7 秒

    關係：依序執行

    結果：完成

  9. 7. 雙語翻譯

    工具：vnt-sentence（句子雙語翻譯）

    處理：將句子翻譯成繁體中文

    產出：sentence_bilingual_transcript.json（雙語逐字稿）

    耗時：28.4 秒

    關係：依序執行

    結果：完成

  10. 8. 雙語時間軸與學習單元

    工具：videonote_learning_timeline_v2.py（建立雙語時間軸與學習單元）

    處理：建立雙語時間軸與學習內容

    產出：sentence_bilingual_timeline.html（雙語時間軸頁）、learning.json（學習單元）、japanese_learning.json（日文學習單元）

    耗時：107.2 秒

    關係：依序執行

    結果：完成

  11. 9. 生成筆記摘要

    工具：videonote-generate-note（生成筆記摘要）

    處理：生成筆記摘要與重點

    產出：note.md（筆記 Markdown）、note.html（筆記 HTML）

    耗時：77.5 秒

    關係：依序執行

    結果：完成

  12. 10. 建立筆記頁面

    工具：videonote-render-html（產生筆記 HTML）

    處理：產生可閱讀的筆記 HTML 頁面

    產出：note.html（筆記 HTML）

    耗時：0.5 秒

    關係：依序執行

    結果：完成

  13. 11. 發布公開頁面

    工具：videonote-publish（發布公開頁面）

    處理：發布到公開網站與每日列表

    產出：公開筆記網址、result.json（管線結果）

    耗時：0.6 秒

    關係：依序執行

    結果：完成

手動來源字幕

自動來源字幕

Whisper 字幕

Qwen 字幕

Parakeet 字幕

MOSS 字幕

聲學時間證據（PyAnnote / WhisperX）

校正前後逐字稿

目前待解決問題與初步改善方向

進一步精進字幕與筆記
使用 Workbench 上傳校正字幕、中文參考稿或回饋 SRT，重新產生雙語與學習單元。
開啟 Workbench