wendy7756/AI-Video-Transcriber

What it solves

AI Video Transcriber provides a unified way to convert video and audio content from over 30 platforms (like YouTube, TikTok, and Bilibili) or local files into structured text. It eliminates the need to manually transcribe audio or hunt for subtitles, while adding AI-driven summarization and translation to make long-form content more digestible.

How it works

The tool uses a multi-stage pipeline to process media:

1. Extraction: It first attempts to extract native subtitles from the platform. If none are found, it uses Faster-Whisper to transcribe the audio locally.
2. Processing: For local uploads, it uses FFmpeg to normalize audio to a standard format for Whisper.
3. AI Enhancement: The resulting transcript is sent to an OpenAI-compatible LLM to correct typos, complete sentences, and organize the text into paragraphs.
4. Finalization: The system generates a summary in one of 11 supported languages and translates the transcript if the summary language differs from the source.

Who it’s for

Content creators, researchers, and students who need to archive, summarize, or translate video and audio content from the web or local storage.

Highlights

• Subtitle-First Architecture: Prioritizes native subtitles for near-instant results, using Whisper only as a fallback.
• Broad Platform Support: Compatible with any site supported by yt-dlp (YouTube, TikTok, Bilibili, etc.).
• Flexible AI Integration: Supports any OpenAI-compatible API endpoint, allowing users to use OpenAI, OpenRouter, or local LLMs.
• Agent-Ready: Includes a headless CLI, a Codex plugin, and an MCP server for integration with AI agents like Claude Code.
• Local File Support: Handles various media formats (.mp3, .mp4, .wav, etc.) and plain text files.