Guangzhou, China · 2026 · AI tool
AI watch pipeline
A link shared from my iPhone becomes a note: the video is downloaded, transcribed by Whisper, then sorted by topic. No manual step.
I save dozens of technical videos every week (Instagram, TikTok, YouTube). Watching them one by one did not scale.
An iPhone shortcut drops the link into a synced file. A scheduled task on my PC picks it up, downloads the video, extracts the audio with ffmpeg, transcribes it locally with Whisper and writes a Markdown note with the transcript and three key frames. Notes are then analysed and grouped by topic.
The code adapts when platforms change: in October 2026 Instagram dropped its separate audio tracks, so the pipeline now takes the sound from the full video.