Transcribee 🐝

**Quick verdict:** Transcribee 🐝 is a focused, practical skill for turning YouTube videos and local media into structured transcripts with speaker diarization. It is well-suited for research, podcast workflows, and downstream LLM analysis, provided you are comfortable with CLI tools and cloud transcription services. **What it does well** - Handles **YouTube URLs and local media** (audio/video) using `yt-dlp` and `ffmpeg`, covering most real-world content sources. - Produces **multiple output artifacts**: - `transcription.txt` – main transcript (likely clean and organized). - `transcription-raw.txt` – plain text without speakers. - `transcription-raw.json` – word-level timings and rich metadata. - `metadata.json` – video info, language, category. - Includes **speaker diarization**, which is valuable for multi-speaker content (podcasts, interviews, meetings). - Uses **ElevenLabs** for transcription, which generally implies strong ASR quality for many languages and accents. **Key risks & limitations** - **External dependencies**: Requires `yt-dlp` and `ffmpeg` (`brew install yt-dlp ffmpeg`), which may limit ease-of-use on non-macOS systems or less technical users. - **Cloud processing & privacy**: Media is sent to ElevenLabs for transcription. This may be unsuitable for **sensitive or confidential recordings** unless you’ve evaluated their privacy/ToS and have appropriate permissions. - **YouTube ToS considerations**: Automated downloading/transcribing of YouTube content may be restricted by YouTube’s terms of service, especially for copyrighted material without rights/permissions. - **Cost & rate limits**: ElevenLabs APIs are typically **paid and rate-limited**, so very long or large-scale transcription workloads may incur non-trivial costs. - **Diarization accuracy**: Speaker separation quality will vary by audio conditions (overlapping speech, background noise), and diarization labels may still require manual cleanup for professional use. **Recommended scenarios** - Researchers, students, or knowledge workers who want to **transcribe and analyze YouTube lectures, talks, and interviews** with LLMs. - Podcasters and content creators who need **speaker-labeled transcripts** for editing, show notes, and content repurposing. - Teams with recorded meetings or interviews who are comfortable sending media to a cloud ASR provider and want **structured outputs** (timings + metadata) for further automation. **Less ideal scenarios** - Highly confidential or regulated audio where third-party cloud processing is not allowed. - Users who require a **purely GUI-based** or fully cross-platform solution without installing CLI dependencies. - Live or streaming transcription (this tool is built around recorded media, not real-time use).