Behind the Scenes: How We Made 28 YouTube Videos in 30 Minutes 🎬

How a dad, a kid, and an AI turned a pile of raw gameplay footage into two full playlists — without a single hour in a video editor.

We finally have a YouTube presence for Blocks Beyond the Stars — 22 Shorts and 6 longer Let's Play / tutorial videos. And honestly? The way they got made is a story worth telling, because the actual "editing" took about half an hour. Here's the full pipeline.

Step 1: Justus hits record 🎮

Every good gaming video starts with someone actually playing the game — and that job obviously went to Justus. He recorded several gameplay sessions with OBS Studio, commentary included. No script, no retakes, just him playing and talking about what he's doing. That raw, honest energy is exactly what we wanted anyway — nobody needs a 10-year-old reading cue cards. 😄

Step 2: Turning audio into data 📝

Raw footage is great, but an AI can't (efficiently) "watch" hours of video. So the trick was to make the videos readable: I uploaded them to ElevenLabs and had it generate transcriptions with timestamps, as JSON. Suddenly every video wasn't a wall of pixels anymore — it was a searchable document: who said what, when.

Step 3: Let Claude watch the footage 🤖

Then everything — videos plus their timestamped transcripts — went into a Claude Cowork project folder. The prompt was basically: "Look at these videos using the timestamp files and figure out how to cut good YouTube Shorts and longer videos out of them."

I also gave it our CI basics — brand colors, fonts, that kind of thing — so titles and overlays would actually look like our game and not like a random template.

Step 4: Brainstorm first, plan second, THEN execute 🗺️

This is the part I'd underline twice, because it's the same lesson we learned from AI-assisted coding: don't let the AI just start doing things.

First we brainstormed together — which moments are Short material, which are boring, what makes a good hook, what to skip entirely. Then I told Claude to write a detailed plan before touching anything (exactly like we do for code changes). The deliverables were clear:

• the cut videos themselves — Shorts and long-form,

• burned-in English and German subtitles (a bilingual family project stays bilingual, always 🇩🇪🇬🇧),

• and a Markdown file with ready-to-paste YouTube descriptions for every single video.

  
  

Step 5: Voilà ✨

About 30 minutes later: 22 Shorts and 6 longer videos, subtitled in two languages, each with its own description, ready for upload. The most time-consuming part of the whole production was Justus playing the game — which, let's be honest, doesn't count as work in his book. 😎

Step 6: The plot twist — uploading took longer than editing 🐢

Here's the funny part: producing 28 videos took 30 minutes. Getting them onto YouTube took 40. Every single video had to be uploaded, described, and scheduled by hand — click, paste, set the release date, repeat, 28 times. There's simply no API for this part of the workflow, so no amount of AI magic could save us. The most futuristic video pipeline we've ever built, bottlenecked by copy & paste. 🙃

What we'd tell anyone trying this

The magic ingredient isn't any single tool — it's the timestamps. Once your footage exists as structured text, an AI can reason about pacing, hooks and highlights just like it reasons about code. Plan first, execute second. Same workflow, new medium. And budget more time for the upload clicking than for the "editing" — 2026 is a strange place. 😄

📺 Watch the results

• Shorts playlist: Shorts

• Let's Plays & tutorials: Let's Play & Tutorials

  
  

Like, subscribe, and tell us which moment deserves to be the next Short. 🚀