工具 API

video_transcript_submit

Submit a video URL to extract its transcript (caption / spoken text). Returns a job_id immediately; poll with video_transcript_result. Supports X / Twitter video links only: - https://x.com/<user>/status/<id>/video/1 - https://twitter.com/<user>/status/<id> Pipeline (auto fallback): 1. Pull video's native auto-subtitle via yt-dlp (free, fastest) 2. OpenRouter Whisper (requires OPENROUTER_API_KEY; default model openai/whisper-large-v3-turbo — full punctuation, near-ElevenLabs quality at ~10× lower cost; override via VIDEO_TRANSCRIPT_OPENROUTER_MODEL) 3. Alibaba DashScope Paraformer (requires DASHSCOPE_API_KEY, ¥0.22/min, China-friendly, fallback when OpenRouter fails) 4. Local faster-whisper fallback (if installed) Use cases / keywords: X video transcript, Twitter video transcript, tweet video caption, X video to text, Twitter video to text, 推文视频转录, X 平台视频文案提取, Twitter 视频字幕转写, 推文视频字幕, 视频内容二次创作。 Args: url: X/Twitter video URL, e.g. "https://x.com/TheRundownAI/status/2071676839152452019/video/1", language: Optional language hint (BCP-47 short code like "en", "zh", "ja"). If omitted, Whisper will auto-detect. Returns: JSON with job_id, status=pending, platform, video_id. Poll with video_transcript_result.