videosays

The problem

Transcribing video content manually is slow and error-prone. Building a reliable scraper and speech-to-text pipeline for fragmented social media platforms like TikTok, Douyin, and YouTube is a maintenance nightmare.

What it does

• Extracts spoken text from YouTube, TikTok, Douyin, and Xiaohongshu links.
• Provides plain text transcripts, timestamped timelines, or SRT/VTT subtitle files.
• Handles asynchronous processing with built-in task status monitoring.
• Manages authentication via browser-based OAuth or direct API key injection.
• Tracks credit balance and transcription history directly from the terminal.

Frameworks & tools

Built as a Node.js CLI tool. Requires npx and Node.js 18 or higher.

Why this beats prompting it yourself

Standard LLMs cannot visit video URLs or "watch" video files directly. This skill handles the heavy lifting of media fetching and speech-to-text processing, feeding clean text back into your agent's context window.

Use cases

• Generating summaries for long-form YouTube tutorials or interviews.
• Extracting captions from TikTok or Douyin clips for content repurposing.
• Creating subtitle files (SRT/VTT) for video production workflows.
• Monitoring credit usage across multiple transcription tasks.

Known limitations

Requires an active internet connection and a Videosays account. Very long videos may return a pending status requiring polling via the task ID.