Give Claude Code the power to watch videos frame by frame. Follow these steps and you can have it running in under 10 minutes.
This setup wires Google's Gemini 2.5 Flash into Claude Code. After that, Claude can actually watch a clip second by second. It tells you what is on screen and what each shot looks like. Drop in a video and Claude returns a timestamped breakdown of what is happening. That is Claude watching the footage, not just reading a transcript.
Creators and researchers who want Claude to break down videos instead of scrubbing and rewinding.
Gemini is Google's model and the video calls it one of the best in the world at understanding video. You need an API key so Claude Code can call it. Grab the key once and keep it somewhere safe. Never share it with anyone.
The video says signing in is free and generating the key is free. Google AI Studio gives you a free tier for testing and getting started.
Anyone who wants Claude to understand real video, not just a transcript.
The video handles this up front because it is the most common question. The free tier covers casual use, testing, and getting started. When you move to paid, the video calls Gemini 2.5 Flash one of the cheapest models out there. You pay fractions of a cent per request. Breaking down a short clip costs less than a penny. Run it on dozens of videos a week and you are looking at a few dollars a month at most. For what it does, the cost is very low next to the time you save.
People who worry about runaway bills before they even try a new tool.
Open Claude Code inside your project folder. Paste one prompt and Claude does the rest. In the video it installs the libraries, writes the Python script, and shows you how to run it. You do not touch a single command yourself.
I want to add Gemini 2.5 Flash video analysis to my Claude Code setup. Please:
1. Create a .env file in this folder and add a placeholder: GEMINI_API_KEY=your_key_here
2. Install the required libraries: google-genai and python-dotenv
3. Write a Python script called analyse_video.py that loads the API key from .env, sends a video file to Gemini 2.5 Flash, and returns a description of what's happening in the video
4. Show me how to run it
Beginners who would rather paste one prompt than run installs by hand.
Claude creates a .env file with a placeholder key. Swap that placeholder for your real key so the script can call Gemini. Open the .env file, paste your key, and save it. Claude refreshes and picks up the new key. Now you are ready to run.
Anyone setting up their first .env file and keys.
Time to prove it works. Drop a short video file into your folder. Tell Claude to run the script against it. Gemini watches the clip and returns a second-by-second breakdown. In the video you see a line like "at 0:01, a logo on a skyscraper at night." Jump to that timestamp. It matches. That is Claude watching, not guessing.
Creators who want proof the tool actually sees what is on screen.
This is just the setup. The video teases what you stack on top next. It is a full reframe analyzer. You drop in any short video URL. It comes back with a complete shot-by-shot breakdown in a clean HTML report. That is where this gets powerful for creators. Start with this base, then build up.
Creators ready to turn this base into a full research workflow.
🚀 Want the full walkthrough on video? It is on the channel: youtube.com/@alisonbuilds