From Documentary to Infographic: An AI Adventure

I recently watched an Arte TV documentary about the fall of ancient Egypt - Der Untergang des alten Ägypten. It was fascinating, packed with information, and naturally I forgot half of it by the next morning. That got me thinking: could I turn a video documentary into a graphical overview that captures the key information in a way that is easy to revisit?

Turns out, with a few tools and some AI, you absolutely can.

The pipeline

The idea is simple: get the spoken content out of the video, transcribe it, and then use AI to generate a structured visual summary. Here is how I did it step by step.

Step 1: Download the video. Arte makes its documentaries available in their media library, but downloading them directly is not straightforward. I used MediathekViewWeb to find and download the video file.

Step 2: Extract the audio. I did not need the video itself, just the spoken content. A quick ffmpeg command strips the audio track from the video file:

ffmpeg -i documentary.mp4 -vn -acodec aac output.aac

Step 3: Transcribe with Whisper. OpenAI's Whisper model does an impressive job at speech-to-text, even with German audio. You can install it locally and run the transcription with just two commands:

pip install -U openai-whisper
whisper output.aac --model base

The result was a surprisingly accurate transcript of the full documentary.

Step 4: Generate the infographic. This is where it gets interesting. I fed the transcript into Claude and asked it to create a graphical info sheet summarizing the documentary's content. Claude generated a structured, visual HTML page covering the key events, timelines, and relationships described in the documentary.

The result is available at content.oglimmer.com/s/egypt if you want to see it yourself.

What makes this fascinating

The entire pipeline - from a two-hour documentary to a shareable visual summary - took maybe 30 minutes of hands-on work. Most of that was waiting for the transcription to finish. The actual creative and analytical heavy lifting was done by AI.

What surprised me most was the quality. The transcript captured the narration accurately, and Claude did a remarkable job at extracting the important themes, organizing them visually, and presenting them in a way that genuinely helps you understand and remember the content.

Is this the end of school books?

This experiment raises a bigger question. If anyone can take a documentary, a lecture, or any other educational content and turn it into personalized learning material in minutes - what does that mean for traditional educational resources?

Today, students use textbooks that were written years ago, designed for a generic audience, and updated on slow publication cycles. But the tools now exist for students to create their own learning material. Watch a documentary, attend a lecture, read a paper - then use AI to generate summaries, infographics, flashcards, or whatever format works best for your own learning style.

The material is not just consumed anymore. It is transformed, personalized, and made your own. That is a fundamentally different relationship with educational content.

I am not saying school books will disappear tomorrow. But the direction is clear: the future of learning material is personal, AI-assisted, and created on demand. The internet provides the raw content. AI provides the transformation. The student decides the format.

That is a fascinating shift, and we are just at the beginning of it.