mediasense

Give Claude ears and eyes for video and audio

Claude can't open a TikTok, a YouTube video or an Instagram reel on its own. mediasense is a free, open-source MCP server that lets it: paste a link (or point at a file) in Claude Desktop or Claude Code and ask in plain language. Claude gets the transcript with a timestamp for every word, the key frames of each shot, and tools to cut and edit. Everything runs on your computer.

Install (1 command) GitHub PyPI

Claude summarizing a TikTok video with mediasense: download, transcription and a summary with timestamps

Real output for a 17-second TikTok by @elim.doc, shown with the creator's permission.

What you can ask Claude

summarize this https://www.tiktok.com/@someone/video/7350…
what does he show at 0:30? and what tool is he talking about?
clean up ~/Desktop/podcast.m4a — remove the "um"s and long pauses
make short captions for ~/Videos/reel.mp4
put intro.mp3 under the first 5 seconds of talk.mp4, ducked

Features

Summarize videos	TikTok, YouTube, Instagram, X, Vimeo and ~1,800 other sites, or any local audio/video file.
Transcribe	About 99 languages with faster-whisper, a timestamp for every word, voice-activity filtering and vocabulary hints for names and jargon.
See the video	Key frames at scene changes and labelled contact sheets, so Claude can describe what is on screen.
Edit by text	Remove filler words, pauses or any sentence; Claude shows the cut list before rendering a new file. The original is never touched.
Subtitles	Export SRT, VTT, plain text, EDL or FCPXML for Premiere, DaVinci Resolve and Final Cut.
Place audio	Put music, a voice-over or a sound effect into a video at the right moment, with ducking.
Private	No account, no API key, no telemetry. Media and transcripts stay on your machine (privacy policy).
Fast on any PC	Works on CPU; NVIDIA GPU acceleration is an optional extra.

Install

Windows (PowerShell):

irm https://raw.githubusercontent.com/joaoestella/mediasense/main/install.ps1 | iex

macOS / Linux:

curl -LsSf https://raw.githubusercontent.com/joaoestella/mediasense/main/install.sh | sh

Then quit Claude Desktop completely, open it again and ask "summarize this video: <link>". In Claude Code you can also install it as a plugin:

/plugin install mediasense --marketplace joaoestella/mediasense

Any other MCP client: uvx mediasense mcp (stdio).

Command line

mediasense understand https://youtu.be/…      # transcript + shots + key frames
mediasense transcribe talk.mp4 --language pt
mediasense edit talk.mp4 --fillers --silences 0.6 --srt
mediasense export talk.transcript.json talk.srt talk.vtt